Project data whole cycle tracing method and system based on multi-source data fusion

Abstract

This application relates to the field of data processing technology, specifically to a method and system for full-cycle traceability of project data based on multi-source data fusion, solving the technical problem of difficulty in achieving accurate monitoring and full-cycle data traceability of the health status of engineering structures in existing technologies. The method includes: collecting stress time-series data and displacement time-series data of key stress-bearing parts of the engineering structure; identifying abnormal monitoring points based on the degree of deviation of the time-series characteristics of each monitoring point and the abnormal synchronous correlation characteristics between monitoring points and other monitoring points; adaptively filtering the stress time-series data of each monitoring point to obtain filtered stress time-series data; fusing the filtered stress time-series data with displacement time-series data, engineering project design data, construction data, and environmental monitoring data to obtain feature-fused data; and establishing a data traceability link based on the feature-fused data to characterize the evolution of the structural stress state throughout the entire lifecycle of the engineering project.

Description

Technical Field

[0001] This application relates to the field of data processing technology, specifically to a method and system for full-cycle traceability of project data based on multi-source data fusion. Background Technology

[0002] The entire lifecycle management of large-scale building and infrastructure projects involves multiple stages, including design, construction, acceptance, and operation and maintenance. Each stage generates a large amount of heterogeneous data, including structural monitoring data, design drawings, construction logs, and environmental parameters. With the development of the Internet of Things and sensing technology, stress sensors, strain gauges, and displacement gauges can be deployed at key stress-bearing parts of engineering structures to achieve real-time monitoring of the structural stress state, acquire stress time-series data and displacement time-series data, and, together with engineering design data, construction records, and environmental monitoring data, provide a multi-dimensional data foundation for structural safety assessment.

[0003] In existing technologies, structural health monitoring typically employs fixed threshold methods or simple filtering to process monitoring data, and stores and manages this data separately from data from design and construction phases. This approach fails to fully consider the spatiotemporal correlations between monitoring data and the dynamic evolution of structural risk conditions, and it also fails to achieve effective fusion and cross-phase correlation of multi-source heterogeneous data, resulting in a disconnect between data processing and project management. Therefore, existing technologies struggle to achieve accurate monitoring of the structural health status and full-cycle data traceability. Summary of the Invention

[0004] To address the technical challenges of accurately monitoring the health status of engineering structures and tracing data throughout their entire lifecycle in existing technologies, this application aims to provide a method for tracing project data throughout its entire lifecycle based on multi-source data fusion. The specific technical solution adopted is as follows: Collect stress time series data and displacement time series data of key stress-bearing parts of the engineering structure; Based on the degree of deviation of the temporal characteristics of each monitoring point and the abnormal synchronous correlation characteristics between the monitoring points and other monitoring points, abnormal monitoring points are identified at each monitoring time. Based on the spatial distribution concentration of abnormal monitoring points and the overall structural risk evolution trend, the stress time series data of each monitoring point are adaptively filtered to obtain filtered stress time series data; the overall structural risk evolution trend is used to characterize the degree of expansion or contraction of the structural abnormal risk area over time. Feature fusion is performed on the filtered stress time series data, displacement time series data, engineering project design data, construction data, and environmental monitoring data to obtain feature fused data; Based on feature fusion data, a data traceability link is established to characterize the evolution of structural stress state throughout the entire life cycle of an engineering project.

[0005] In one possible implementation, abnormal monitoring points are identified at each monitoring time based on the degree of deviation of the temporal characteristics of each monitoring point and the abnormal synchronous correlation characteristics between the monitoring points and other monitoring points. This includes: for each monitoring time, selecting monitoring points that meet preset abnormality conditions from all monitoring points based on the degree of deviation of the temporal characteristics of each monitoring point at the monitoring time, and determining them as temporal abnormality candidate points; the degree of deviation of the temporal characteristics is used to characterize the deviation of the monitoring point from historical trend data at the current monitoring time; and determining abnormal monitoring points from the temporal abnormality candidate points based on the abnormal synchronous correlation characteristics between the temporal abnormality candidate points and other monitoring points; the abnormal synchronous correlation characteristics are used to characterize the degree of coordination between the temporal abnormality candidate points and surrounding monitoring points in the same time period.

[0006] In one possible implementation, the process of determining the degree of deviation of the time series characteristics includes: acquiring historical stress time series data of the monitoring point before the monitoring time; processing the historical stress time series data based on the trend extraction method to obtain trend change data; calculating the residual data of the stress data at the current monitoring time relative to the trend change data; and determining the degree of deviation of the time series characteristics based on the distribution characteristics and trend characteristics of the residual data.

[0007] In one possible implementation, the process of determining the abnormal synchronization correlation features includes: obtaining the sequence of deviations in time-series features between the candidate time-series anomaly points and other monitoring points within the same time period; analyzing the changing trend of the time-series feature deviation sequence and determining the synchronization change parameter based on the similarity of the changing trend; and determining the abnormal synchronization correlation features based on the distance parameter between the candidate time-series anomaly points and other monitoring points, as well as the synchronization change parameter.

[0008] In one possible implementation, adaptive filtering is performed on the stress time series data of each monitoring point based on the spatial distribution concentration of abnormal monitoring points and the overall structural risk evolution trend. This includes: clustering abnormal monitoring points based on spatial distribution characteristics to obtain risk regions; risk regions are used to characterize a set of monitoring points that are spatially continuous and have synchronous stress anomalies; determining the overall structural risk index based on the number of risk regions, the number of monitoring points within each risk region, and the spatial distribution density between risk regions; determining the overall structural risk evolution trend based on the time series variation characteristics of the risk region area; the risk region area is the total number of abnormal monitoring points contained in all risk regions at each monitoring time; determining the attention factor based on the overall structural risk index and the overall structural risk evolution trend; setting differentiated filtering weights for different monitoring points based on the attention factor, and performing adaptive filtering based on the differentiated filtering weights.

[0009] In one possible implementation, the overall structural risk index is determined based on the number of risk areas, the number of monitoring points within each risk area, and the spatial distribution density between risk areas. This includes: determining the total number of risk areas and counting the number of monitoring points within each risk area; determining the average distance between risk areas based on the mean distance between all risk areas, and determining a spatial distribution density parameter based on the average distance; the spatial distribution density parameter characterizes the spatial concentration of risk areas; for each risk area, determining a risk scale parameter based on the number of monitoring points within the risk area and the mean stress anomaly level of all monitoring points within the risk area; summing the risk scale parameters of all risk areas to obtain an overall risk area scale parameter; and determining the overall structural risk index based on the overall risk area scale parameter and the spatial distribution density parameter.

[0010] In one possible implementation, the focus factors are determined based on the overall structural risk index and the overall structural risk evolution trend. This includes: obtaining the risk area area at each monitoring time; the risk area area is the total number of abnormal monitoring points contained within all risk areas at the corresponding monitoring time; determining the change in risk area area based on the difference between the risk area area at the previous monitoring time and the risk area area at the current monitoring time; determining the risk area change trend parameter based on the slope of the trend line fitted from the risk area areas at each monitoring time; determining the overall structural risk evolution trend based on the product of the change in risk area area and the risk area change trend parameter; and normalizing the product of the overall structural risk index and the overall structural risk evolution trend to obtain the focus factors.

[0011] In one possible implementation, feature fusion is performed on the filtered stress time series data, displacement time series data, engineering project design data, construction data, and environmental monitoring data to obtain feature-fused data. This includes: spatiotemporally aligning the filtered stress time series data with the displacement time series data, design data, construction data, and environmental monitoring data; for the spatiotemporally aligned data: extracting stress statistical features from the filtered stress time series data; extracting design benchmark features from the design data, and determining design deviation features based on the differences between the filtered stress time series data and the design benchmark features; extracting construction stage identification features and construction load application features from the construction data; and extracting environmental... Environmental statistical features are extracted from monitoring data; displacement statistical features are extracted from displacement time series data; stress statistical features include mean, standard deviation, peak value, frequency, and fluctuation range; displacement statistical features include displacement change, maximum displacement, and displacement rate features; environmental statistical features include temperature influence features, humidity influence features, wind speed influence features, vibration influence features, and rainfall influence features; based on dimensionality reduction technology, the extracted stress statistical features, displacement statistical features, design deviation features, construction stage identification features, construction load application features, temperature influence features, humidity influence features, and wind speed influence features are fused into a low-dimensional feature vector; the low-dimensional feature vector is used as feature fusion data.

[0012] In one possible implementation, a data traceability link is established based on feature fusion data to characterize the evolution of the structural stress state throughout the entire lifecycle of an engineering project. This includes: associating and storing the feature fusion data with data from the design, construction, acceptance, and operation and maintenance phases at corresponding monitoring times to construct a full-cycle database; adding timestamp tags, spatial location tags, and process phase tags to the feature fusion data and data from each phase to form a multi-dimensional data index; establishing cross-phase data traceability paths based on the multi-dimensional data index; generating a visual traceability map based on the cross-phase data traceability paths; and using the visual traceability map to display the correlation between the structural stress state and design benchmarks, construction activities, and environmental factors at different phases.

[0013] This application also provides a project data full-lifecycle traceability system based on multi-source data fusion, including: The data acquisition unit is used to collect stress time-series data and displacement time-series data of key stress-bearing parts of the engineering structure; An anomaly identification unit is used to identify abnormal monitoring points at each monitoring time based on the degree of deviation of the temporal characteristics of each monitoring point and the abnormal synchronization correlation characteristics between the monitoring points and other monitoring points. The filtering unit is used to adaptively filter the stress time series data of each monitoring point according to the spatial distribution concentration of abnormal monitoring points and the overall risk evolution trend of the structure, so as to obtain the filtered stress time series data; the overall risk evolution trend of the structure is used to characterize the degree of expansion or contraction of the abnormal risk area of ​​the structure over time. The feature fusion unit is used to perform feature fusion on filtered stress time series data, displacement time series data, engineering project design data, construction data and environmental monitoring data to obtain feature fused data. The link establishment unit is used to establish a data traceability link based on feature fusion data to characterize the evolution of the structural stress state throughout the entire life cycle of an engineering project.

[0014] This application has the following beneficial effects: By collecting stress and displacement time-series data from key stress-bearing parts of the engineering structure, this application obtains core basic data that can directly reflect the true stress state of the structure; then, by combining the deviation degree of time-series characteristics of a single monitoring point with the synchronous correlation characteristics of anomalies of multiple monitoring points, it accurately identifies abnormal monitoring points at each monitoring time, effectively distinguishing between real structural anomalies and false interference signals; subsequently, based on the spatial distribution concentration of abnormal monitoring points and the overall risk evolution trend of the structure, it performs adaptive filtering on the stress time-series data, filtering out invalid noise while completely retaining the true structural anomaly signals; then, it performs feature fusion with multi-source data from project design, construction, and environment, breaking down data silos and forming feature fusion data that can comprehensively reflect the structural state and influencing factors throughout the entire cycle; finally, based on the feature fusion data, it establishes a data traceability link that runs through the entire project cycle, realizing accurate traceability of the entire process of structural stress state evolution. Through the above complete technical solution, this embodiment solves the technical problems in the prior art, such as the susceptibility of structural monitoring data to interference, low accuracy of anomaly identification, inability to deeply integrate multi-source data, and difficulty in achieving accurate traceability throughout the entire life cycle. It effectively improves the accuracy of engineering structural safety assessment and realizes refined and intelligent management and control of structural safety throughout the entire life cycle of engineering projects. Attached Figure Description

[0015] To more clearly illustrate the technical solutions and advantages in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0016] Figure 1 A flowchart illustrating a method for full-lifecycle traceability of project data based on multi-source data fusion, provided as an embodiment of this application; Figure 2This is a schematic diagram of the system architecture of a project data full-cycle traceability system based on multi-source data fusion, provided as an embodiment of this application. Detailed Implementation

[0017] To further illustrate the technical means and effects adopted by this application to achieve the intended purpose of the invention, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation, structure, features, and effects of a project data full-lifecycle traceability method based on multi-source data fusion proposed in this application. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0018] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains.

[0019] The following description, in conjunction with the accompanying drawings, details a specific scheme for a project data full-cycle traceability method based on multi-source data fusion provided in this application.

[0020] Please see Figure 1 It illustrates a flowchart of a project data full-lifecycle traceability method based on multi-source data fusion provided in an embodiment of this application, such as... Figure 1 As shown, the method includes the following steps: Step 101: Collect stress time series data and displacement time series data of key stress-bearing parts of the engineering structure.

[0021] Among them, the key stress-bearing components of the engineering structure are the core parts that determine the overall stability and load-bearing capacity of the structure in an engineering project. Examples include beams, columns, walls, and foundation pit support structures in building structures; main beams and piers in bridge engineering; and linings and surrounding rock in tunnel engineering. Stress time-series data are sequence data of stress values ​​at corresponding monitoring points changing over time, collected synchronously at fixed time intervals. Displacement time-series data are sequence data of settlement, displacement, and tilt angle values ​​at corresponding monitoring locations changing over time, collected synchronously with the stress time-series data.

[0022] Specifically, stress sensors, strain gauges, and other data acquisition devices can be installed at key stress-bearing parts of the engineering structure to synchronously collect stress values ​​at each monitoring point at a preset sampling frequency, forming stress time series data. At the same time, displacement gauges, inclinometers, and other devices can be installed at the corresponding monitoring sections to synchronously collect displacement, settlement, and tilt data of the structure at the same sampling frequency, forming displacement time series data that is spatiotemporally aligned with the stress time series data. The collected data can be uploaded to a local data terminal or cloud platform for storage and subsequent processing.

[0023] Step 102: Identify the abnormal monitoring points at each monitoring time based on the degree of deviation of the temporal characteristics of each monitoring point and the abnormal synchronization correlation characteristics between the monitoring points and other monitoring points.

[0024] Among them, the degree of deviation of the time series feature is used to characterize the deviation of the monitoring point from the historical trend data at the current monitoring time, and can reflect whether the stress data of a single monitoring point has an abnormal change. The anomaly synchronization correlation feature is used to characterize the degree of coordination between the monitoring point and other surrounding monitoring points in the same period of time, and can distinguish whether the anomaly of a single monitoring point is due to sensor interference or the overall stress anomaly of the structure.

[0025] Optionally, for each monitoring time, first calculate the degree of deviation of the temporal characteristics of all monitoring points at that time, and screen out candidate monitoring points with temporal anomalies; then calculate the abnormal synchronization correlation characteristics between the candidate monitoring points and other monitoring points, and combine spatial synchronization to finally determine the real abnormal monitoring points at that monitoring time from the candidate monitoring points.

[0026] Step 103: Based on the spatial distribution concentration of abnormal monitoring points and the overall risk evolution trend of the structure, adaptive filtering is performed on the stress time series data of each monitoring point to obtain the filtered stress time series data.

[0027] Among them, the overall structural risk evolution trend is used to characterize the degree of expansion or contraction of structural anomaly risk areas over time, reflecting the dynamic direction of structural safety risk changes. The spatial distribution concentration of anomaly monitoring points is used to characterize the degree of aggregation of anomaly monitoring points in the engineering structure space, reflecting the spatial distribution characteristics of anomaly areas. Adaptive filtering refers to the ability to set differentiated filtering weights for different monitoring points according to the structural risk status, achieving differentiated filtering processing for data of different importance.

[0028] In some embodiments, spatial clustering analysis is first performed on the identified abnormal monitoring points to determine the risk areas and spatial distribution concentration, and the overall structural risk index is calculated; then, based on the temporal changes of the risk areas, the overall structural risk evolution trend is determined; combining the risk index and the evolution trend, differentiated filtering weights are set for different monitoring points, and adaptive filtering is performed on the stress time series data to obtain denoised stress time series data that retains the true structural characteristics.

[0029] In this step, by combining the spatial distribution characteristics of abnormal monitoring points with the risk evolution trend, adaptive filtering of stress time series data is achieved. This can effectively filter out invalid fluctuations caused by environmental noise and sensor interference, while retaining the effective signals corresponding to real structural anomalies. This avoids the excessive smoothing of real anomaly signals by traditional fixed parameter filtering, and ensures the accuracy of subsequent data fusion and tracing.

[0030] Step 104: Perform feature fusion on the filtered stress time series data, displacement time series data, engineering project design data, construction data, and environmental monitoring data to obtain feature fused data.

[0031] Design data refers to data generated during the design phase of an engineering project to characterize the structural design benchmark. Examples include design drawings, design specifications, Building Information Modeling (BIM), design stress indicators, allowable stress thresholds, and material property parameters. Construction data refers to data generated during the construction phase of an engineering project to characterize the construction process and procedures. Examples include construction logs, procedure reporting records, construction time, construction location, load application records, and pouring and tensioning records. Environmental monitoring data refers to time-series data of environmental parameters at and around the project site, collected synchronously with the monitoring data. Examples include temperature, humidity, wind speed, vibration, and rainfall data.

[0032] In some embodiments, this application first performs spatiotemporal alignment on all multi-source data to ensure that all data are under a unified time and space benchmark; then extracts key feature parameters that can characterize the corresponding dimensional features from different data sources; finally, through dimensionality reduction techniques, the extracted multi-dimensional high-dimensional features are fused into a unified low-dimensional feature vector as feature fusion data.

[0033] This step enables deep feature fusion of the filtered core monitoring data with multi-source data from the design, construction, and environment throughout the project lifecycle. It breaks down data silos between different data sources, links the structural stress state with design benchmarks, construction activities, and environmental influencing factors, and provides a complete feature foundation for subsequent full-cycle data traceability.

[0034] Step 105: Based on feature fusion data, establish a data traceability link to characterize the evolution of the structural stress state throughout the entire life cycle of the engineering project.

[0035] Among them, the data traceability link refers to a complete path that runs through the entire life cycle of engineering project design, construction, acceptance, operation and maintenance, enabling bidirectional traceability and correlation query of structural stress state data and business data at each stage, and enabling full-cycle source tracing and cause analysis of structural stress anomalies.

[0036] In some embodiments, this application first associates and stores the feature fusion data with the corresponding data of each stage of the project lifecycle to construct a full-cycle database; then adds multi-dimensional index tags to all data to establish a cross-stage data traceability path; finally, based on the traceability path, generates a visualized traceability map to realize full-cycle visualized traceability of the structural stress state evolution.

[0037] Based on the above technical solution, this embodiment acquires core basic data that directly reflects the true stress state of the structure by collecting stress time-series data and displacement time-series data of key stress-bearing parts of the engineering structure. Then, by combining the deviation degree of time-series characteristics of a single monitoring point with the synchronous correlation characteristics of anomalies at multiple monitoring points, it accurately identifies abnormal monitoring points at each monitoring moment, effectively distinguishing between real structural anomalies and false interference signals. Subsequently, based on the spatial distribution concentration of abnormal monitoring points and the overall risk evolution trend of the structure, adaptive filtering is performed on the stress time-series data, filtering out invalid noise while completely preserving the true structural anomaly signals. The filtered monitoring data is then fused with multi-source data from project design, construction, and environment to break down data silos and form feature-fused data that comprehensively reflects the structural state and influencing factors throughout the entire lifecycle. Finally, based on the feature-fused data, a data traceability link is established throughout the entire project lifecycle, achieving precise traceability of the entire process of structural stress state evolution. Through the above complete technical solution, this embodiment solves the technical problems in the prior art, such as the susceptibility of structural monitoring data to interference, low accuracy of anomaly identification, inability to deeply integrate multi-source data, and difficulty in achieving accurate traceability throughout the entire life cycle. It effectively improves the accuracy of engineering structural safety assessment and realizes refined and intelligent management and control of structural safety throughout the entire life cycle of engineering projects.

[0038] In one possible implementation, the above steps identify abnormal monitoring points at each monitoring time based on the degree of deviation of the temporal characteristics of each monitoring point and the abnormal synchronization correlation characteristics between the monitoring points and other monitoring points, including: Step 201: For each monitoring time, based on the degree of deviation of the temporal characteristics of each monitoring point at the monitoring time, select monitoring points that meet the preset abnormal conditions from all monitoring points and determine them as candidate points for temporal anomalies.

[0039] Among them, the degree of deviation of time series features is used to characterize the deviation of the monitoring point from the historical trend data at the current monitoring time.

[0040] Specifically, structural stress data may fluctuate instantaneously due to temperature changes, wind loads, or construction activities, so historical trend analysis is needed to identify the true deviations.

[0041] Taking the i-th monitoring time as an example, for the j-th monitoring point, historical stress time-series data within a preset reference period (exemplarily, the first 60 times) is acquired from the i-th time. This historical data is processed using a trend extraction method (such as moving average filtering) to obtain trend change data, which characterizes the slow change trend of structural stress while smoothing out high-frequency noise and transient spikes. Then, the residual data of the stress data at the current monitoring time relative to the trend change data is calculated, and the degree of deviation of the time-series characteristics is determined based on the distribution characteristics (such as the mean of the absolute values ​​of the residuals) and trend characteristics (such as the fitting slope of the residual sequence). Monitoring points with a time-series characteristic deviation greater than or equal to a preset threshold are identified as candidate points for time-series anomalies. Optionally, this preset threshold is determined based on extensive engineering experience to balance the sensitivity and false alarm rate of anomaly identification; an exemplary value of 0.7 is used.

[0042] Step 202: Based on the abnormal synchronization correlation characteristics between the time-series anomaly candidate points and other monitoring points, determine the anomaly monitoring points from the time-series anomaly candidate points.

[0043] Among them, the abnormal synchronization correlation feature is used to characterize the degree of coordination between the time-series abnormality candidate point and the surrounding monitoring point when stress abrupt changes occur within the same time period.

[0044] Since real structural anomalies have spatial transmission properties, if multiple adjacent monitoring points exhibit similar stress change patterns within the same time period, the anomaly is more likely to originate from the structure's own response rather than local disturbances.

[0045] Therefore, for each candidate point of temporal anomaly, the sequence of its deviation from the temporal characteristics of other surrounding monitoring points within the same time period is obtained. The similarity of the changing trends of these sequences is analyzed to determine the synchronous change parameters. Combined with the distance parameters between monitoring points, the synchronous correlation characteristics of the anomaly are determined.

[0046] For example, this application ultimately determines the candidate points of temporal anomalies whose abnormal synchronization correlation characteristics are greater than or equal to a preset synchronization threshold as anomaly monitoring points. Through this mechanism, isolated anomalies caused by local noise or sensor failures are effectively excluded, ensuring that the identified anomaly monitoring points have real structural spatial correlation.

[0047] Optionally, the preset synchronization threshold is determined comprehensively based on the structural dynamic response characteristics and the level of environmental disturbance: for flexible structures with large wind loads or vibration disturbances (such as tall towers), the threshold can be set to 0.75~0.85 to exclude false synchronizations caused by the environment; for rigid structures (such as dams and tunnels), the threshold can be set to 0.6~0.75. Specifically, the confusion matrix method can be used to calculate the precision and recall at different thresholds on the calibration dataset, and select the threshold when the harmonic mean of precision and recall (F1-score) is maximized, or determine the upper limit of the threshold based on the acceptable false alarm rate for the project (such as <5%), with an example value of 0.7.

[0048] Based on the above technical solution, this application achieves effective differentiation between isolated noise interference and real structural anomalies through a two-level identification mechanism that first screens candidate points of temporal anomalies and then verifies spatial synchronous correlation. By introducing anomaly synchronous correlation features, the spatial propagation characteristics of structural anomalies are fully utilized to ensure that the identified anomaly monitoring points have spatial continuity and synergy, which significantly improves the accuracy and reliability of anomaly identification and provides accurate target positioning for subsequent adaptive filtering.

[0049] As one possible implementation, the process for determining the degree of deviation of the aforementioned time-series characteristics includes: Step 301: Obtain historical stress time series data of the monitoring point before the monitoring time.

[0050] Optionally, taking the j-th monitoring point at the i-th monitoring time as an example, obtain the stress time series data of that monitoring point within a time period consisting of a preset reference period (exemplarily the first 60 times) traced back from the i-th time. This dataset contains all historical observations of the monitoring points from the initial time to the current time.

[0051] Step 302: Process the historical stress time series data based on the trend extraction method to obtain trend change data.

[0052] Optionally, a moving average filtering method can be used to process the stress time series data within the reference period to obtain the trend curve for that period. As a low-pass filtering method, the moving average can effectively preserve the long-term trend of slowly changing structural stress, while having a good smoothing effect on high-frequency noise and transient spikes.

[0053] For example, the size of the moving average window can be set according to the data sampling frequency, such as taking the window width for 5-10 time points.

[0054] Step 303: Calculate the residual data of the stress data at the current monitoring time relative to the trend change data.

[0055] Optionally, the smoothed trend value at the corresponding time point is subtracted from each original stress data point within the reference time period to obtain a residual sequence. This residual sequence characterizes the fluctuation component of the monitoring data that deviates from the normal trend. The larger the residual, the more severe the deviation of the stress value at the monitoring point from the normal trend, and the more likely there is a sudden change.

[0056] Step 304: Determine the degree of deviation of time series characteristics based on the distribution and trend characteristics of the residual data.

[0057] In some embodiments, the mean of the absolute values ​​of the residual sequence corresponding to the i-th monitoring time is calculated. This parameter characterizes the concentration of the residual distribution and reflects the overall magnitude of the deviation trend of the monitoring points. The residual sequence is linearly fitted using the least squares method to obtain the absolute value of the slope of the fitted curve. This parameter characterizes the trend characteristics of the residual changes and reflects the evolution speed of the deviation trend of the monitoring points. Based on the above parameters, a mutation factor is calculated, which serves as a quantitative indicator of the degree of deviation of the time series characteristics.

[0058] For example, mutation factors The following calculation formula must be satisfied:

[0059] in, This represents the mutation factor of the j-th monitoring point at the i-th monitoring time. This represents the mean of the absolute values ​​of the residual sequence corresponding to the i-th monitoring time. This represents the absolute value of the slope of the residual sequence fitting curve corresponding to the i-th monitoring time. () represents the Min-Max normalization function, where the maximum and minimum values ​​can be determined based on historical data.

[0060] This formula comprehensively quantifies the temporal deviation of monitoring points by integrating the dispersion of the residual distribution (characterizing the magnitude of deviation) and the trend of residual changes (characterizing the rate of deviation). When the mean residual is large and shows a continuous increasing trend, the mutation factor value is high, indicating that there is a significant abnormal deviation at the monitoring point.

[0061] Based on the above technical solution, this embodiment effectively separates the long-term trend of structural stress from short-term noise through moving average trend extraction, accurately quantifies the deviation of monitoring points from historical trends through residual analysis, and constructs a mutation factor by fusing residual distribution characteristics and trend characteristics, thereby realizing a multi-dimensional quantitative evaluation of time-series anomalies and providing an objective and accurate measurement standard for screening candidate points of time-series anomalies.

[0062] As one possible implementation, the process for determining the aforementioned abnormal synchronization association characteristics includes: Step 401: Obtain the sequence of deviations in temporal characteristics between the candidate points of temporal anomalies and other monitoring points within the same time period.

[0063] Specifically, for the j-th temporal anomaly candidate point and the g-th other monitoring point (usually spatially adjacent monitoring points, such as monitoring points within four or eight neighborhoods), the mutation period of both is obtained within a reference time period (such as the first 60 times) at the i-th monitoring time. The mutation period consists of consecutive mutation times, and the mutation time refers to the time when the mutation factor of the monitoring point is greater than or equal to a preset threshold (0.7 for example).

[0064] Determine several intersections of the abrupt change periods between two monitoring points. Taking the 0th intersection as an example, this intersection represents the time interval during which both monitoring points simultaneously exhibit anomalies within the same time period. Extract the mutation factor sequence of the j-th monitoring point at all times within this intersection. Similarly, extract the mutation factor sequence at the g-th monitoring point. }

[0065] Step 402: Analyze the changing trend of the time series feature deviation sequence, and determine the synchronous change parameters based on the similarity of the changing trends.

[0066] The two mutation factor sequences were fitted using the least squares method to obtain their respective fitting curves and the slope of the tangent line at each time step. and When the tangent slopes of two monitoring points are simultaneously positive or simultaneously negative at the same moment, it indicates that they are changing in the same direction at that moment, and this moment is recorded as the synchronization moment. Calculate the absolute value of the difference between the two tangent slopes at the synchronization moment | |, using a negative exponential function (such as Process this difference and record the result as the degree of abrupt synchronization at that synchronization moment. The smaller the slope difference, the higher the degree of mutation synchronization, indicating that the abnormal evolution patterns of the two monitoring points are more similar.

[0067] Step 403: Determine the anomaly synchronization correlation characteristics based on the distance parameters between the candidate time-series anomalies and other monitoring points, as well as the synchronization change parameters.

[0068] For the j-th monitoring point and the g-th monitoring point at the i-th monitoring time, calculate the weighted average of the mutation synchronization degree and the number of synchronization times within the intersection, and then normalize the result to obtain the mutation synchronization degree. .

[0069] For example, the degree of mutation synchronization The following calculation formula must be satisfied:

[0070] in, This indicates the degree of synchronization between the mutations at monitoring point j and monitoring point g at monitoring time i; O represents the number of intersections of mutation periods between monitoring point j and monitoring point g at monitoring time i. This represents the number of synchronization moments between the j-th monitoring point and the g-th monitoring point in the o-th intersection. This represents the length (number of moments) of the 0th intersection. This represents the mutation synchronization rate between two monitoring points in the o-th intersection. This represents the maximum value of the average mutation factor of the two monitoring points in the o-th intersection at all synchronization times; () represents the Min-Max normalization function, where the maximum and minimum values ​​can be determined based on historical data. In the above formula, if the denominator... If the calculation result is 0 (i.e., there is no time interval in the intersection), then directly use... The value is assigned to 0. When O=0, there is no synchronous change period between the two monitoring points, and the value is directly set to 0. The value is assigned to 0.

[0071] This formula comprehensively quantifies the abnormal synchronization correlation characteristics between two monitoring points by integrating mutation synchronization rate (temporal synergy) and mutation intensity (abnormal severity).

[0072] After calculating the mutation synchronization degree, monitoring points whose mutation synchronization degree with the j-th monitoring point is greater than or equal to the mutation synchronization degree threshold are designated as reference monitoring points for the j-th monitoring point. Finally, the stress anomaly degree is calculated based on the mutation synchronization degree and spatial distance between the j-th monitoring point and all reference monitoring points. This serves as a comprehensive reflection of the synchronous correlation characteristics of anomalies. It should be noted that the reference monitoring point for each monitoring point does not include the monitoring point itself; for example, the reference monitoring point for the j-th monitoring point does not include the j-th monitoring point. Optionally, the threshold for the degree of synchronization of mutations is determined based on the spatial distribution density and structural continuity characteristics of the monitoring points. For example, for a densely deployed monitoring network (spacing < 5m), the threshold can be 0.6~0.7 to ensure sufficient capture of spatial correlation; for a sparsely deployed monitoring network (spacing > 20m), the threshold can be 0.8~0.9 to avoid misjudgment. Specifically, based on the statistical distribution of synchronous mutation monitoring points in historical real anomaly events, the 85th percentile of the degree of synchronization can be taken as the threshold to ensure that 85% of real anomaly events are identified as having spatial correlation; an example value of 0.7 is used.

[0073] For example, the degree of stress anomaly The following calculation formula must be satisfied:

[0074] in, This indicates the degree of stress anomaly at the j-th monitoring point at the i-th monitoring time; This represents the number of reference monitoring points for the j-th monitoring point at the i-th monitoring time. This indicates the degree of synchronization between the mutations at the j-th monitoring point and the e-th reference monitoring point; This represents the normalized value of the spatial distance (e.g., the Euclidean spatial distance between monitoring points) between the j-th monitoring point and the e-th reference monitoring point. Since the reference monitoring point for each monitoring point does not include the monitoring point itself, therefore... The value of E is always greater than 0. When E=0, the monitoring point is determined to be an isolated anomaly, not a true structural anomaly, and is directly... The value is assigned to 0.

[0075] Optionally, the stress anomaly threshold can be set in stages according to the structural safety level and monitoring sensitivity requirements: for important structural parts of Level I safety (such as the main span of a large bridge), the threshold can be 0.5~0.6 to improve sensitivity; for general structural parts of Level II safety, the threshold can be 0.7~0.8. Specifically, it can be determined by the stress anomaly distribution of historical anomaly samples, taking the mean of stress anomaly under normal working conditions plus 3 times the standard deviation (μ+3σ) as the threshold to ensure that normal fluctuations and true anomalies can be distinguished at a 99.7% confidence level.

[0076] Based on the above technical solution, this embodiment accurately quantifies the degree of anomaly coordination among monitoring points by analyzing the intersection and similarity of change trends during abrupt change periods. By introducing spatial distance weights, the contribution of synchronous anomalies of neighboring monitoring points to the degree of stress anomaly is strengthened, which is consistent with the spatial correlation principle of stress transmission in structural mechanics. By constructing the degree of stress anomaly as a quantitative indicator of anomaly synchronous correlation characteristics, the true structural anomalies (with spatial correlation) and isolated noise interference (without spatial correlation) are effectively distinguished, significantly improving the accuracy and reliability of anomaly monitoring point identification.

[0077] In one possible implementation, the above steps involve adaptive filtering of the stress time series data for each monitoring point based on the spatial distribution concentration of the abnormal monitoring points and the overall risk evolution trend of the structure, including: Step 501: Cluster the abnormal monitoring points based on spatial distribution characteristics to obtain risk areas.

[0078] The risk region is used to characterize a set of monitoring points that are spatially continuous and synchronously stressed.

[0079] Taking all abnormal monitoring points at the i-th monitoring time as the object, and using the spatial distance between any two abnormal monitoring points as the clustering distance, the K-means clustering algorithm is used for clustering to obtain several clusters, each representing a risk region. This risk region corresponds to a continuous, stress-related area on the engineering structure where stress anomalies occur synchronously. During the clustering process, the number of clusters can be adaptively determined based on the number and spatial distribution of abnormal monitoring points, or preset to a fixed value determined based on engineering experience.

[0080] Step 502: Determine the overall structural risk index based on the number of risk areas, the number of monitoring points in each risk area, and the spatial distribution density between risk areas.

[0081] As one possible implementation, this step can be specifically implemented as follows: determine the total number of risk areas, and count the number of monitoring points within each risk area; determine the average distance between risk areas based on the average distance between all risk areas, and determine the spatial distribution density parameter based on the average distance; the spatial distribution density parameter is used to characterize the spatial concentration of risk areas; for each risk area, determine the risk scale parameter of the risk area based on the number of monitoring points within the risk area and the average stress anomaly degree of all monitoring points within the risk area; sum up the risk scale parameters of all risk areas to obtain the overall risk area scale parameter; determine the overall structural risk index based on the overall risk area scale parameter and the spatial distribution density parameter.

[0082] For example, the overall structural risk index The following calculation formula must be satisfied:

[0083] in, This represents the overall structural risk index at the i-th monitoring time. Let represent the mean distance between any two risk areas at the i-th monitoring time. represents the spatial distribution density parameter; F represents the number of risk areas at the i-th monitoring time. This represents the number of monitoring points within the f-th risk area; This represents the average stress anomaly level at all monitoring points within the f-th risk area. For the above risk scale parameters, This refers to the overall scale parameter of the aforementioned risk areas. In the above formula, when there is only one risk area (i.e., F=1), there is no distance between areas, and in this case, we can... The spatial distribution density parameter is set to a preset maximum value (such as the maximum spatial scale of the engineering structure) to approach a minimum value. This formula comprehensively quantifies the overall risk level of the structure by integrating the overall scale of the risk area (number and severity of anomalies) and the spatial distribution density (degree of regional concentration). The larger the scale and the denser the distribution of the risk area, the higher the overall structural risk index, indicating a greater possibility of overall structural hidden dangers. () represents the Min-Max normalization function, where the maximum and minimum values ​​can be determined based on historical data.

[0084] Step 503: Determine the overall structural risk evolution trend based on the temporal variation characteristics of the risk area.

[0085] The risk area is the total number of abnormal monitoring points contained within all risk areas at each monitoring time.

[0086] As one possible implementation, this step can be specifically implemented as follows: obtaining the risk area area at each monitoring time; the risk area area is the total number of abnormal monitoring points contained in all risk areas at the corresponding monitoring time; determining the change in risk area area based on the difference between the risk area area at the previous monitoring time and the risk area area at the current monitoring time; determining the risk area change trend parameter based on the slope of the trend line fitted from the risk area areas at each monitoring time; and determining the overall structural risk evolution trend based on the product of the risk area change and the normalized value of the risk area change trend parameter. For example, this application uses the risk area area sequence of the most recent 50 monitoring times for least-squares linear fitting.

[0087] Step 504: Determine the factors of concern based on the overall structural risk index and the overall structural risk evolution trend.

[0088] Optionally, this step can be implemented as follows: normalize the product of the overall structural risk index and the overall structural risk evolution trend to obtain the attention factor.

[0089] For example, the attention factor satisfies the following calculation formula:

[0090] in, This represents the attention factor at the i-th monitoring time. This indicates the overall structural risk index; This indicates the change in the area of ​​the risk zone; Parameters indicating the trend of changes in the area of ​​the risk zone. and All are dimensionless parameters after normalization. () represents the Min-Max normalization function, which maps the product result to the interval [0,1]. The maximum and minimum values ​​can be determined based on historical data.

[0091] This formula comprehensively assesses the urgency of risk at the current moment by integrating the overall structural risk index (static risk level) and the risk evolution trend (dynamic deterioration rate). The higher the value of the focus factor, the greater the structural risk at the current moment and the worsening trend, the more important the stress data at that moment, and the less filtering is needed to retain detailed features.

[0092] Step 505: Set differentiated filtering weights for different monitoring points based on the factors of interest, and perform adaptive filtering based on the differentiated filtering weights.

[0093] Optionally, the filter weight is the weight ratio of the data point at the current monitoring time in the weighted moving average filter window. The weight value is negatively correlated with the smoothing effect: the larger the weight of the current data point, the higher its proportion in the filtering result, the weaker the smoothing effect, and the more complete the original data details are preserved; the smaller the weight of the current data point, the lower its proportion in the filtering result, the stronger the smoothing effect, and the more fully environmental noise and interference signals are filtered out.

[0094] Specifically, based on the attention factors The size of the filter weight is set differently for different monitoring points: For the abnormal monitoring point at the i-th monitoring time (the monitoring point located in the risk area), since it contains key structural anomaly information, the filter weight is set to the preset maximum weight value to fully preserve the original data details without additional smoothing processing. Optionally, the preset maximum weight value (exemplary value is 10) is determined according to the filter window width and the dynamic range of the data to ensure that the weight of the abnormal monitoring point is 5 to 10 times that of the ordinary monitoring point (minimum weight is about 1), so that the contribution of the abnormal point data in the weighted average exceeds 50%, which is equivalent to almost no averaging processing on the abnormal points. Specifically, it can be determined through simulation test: add noise to the known abnormal signal, and when the weight value makes the attenuation of the abnormal amplitude after filtering less than 5%, this weight is the minimum effective weight. Usually, twice this value is taken as a safety margin, and the exemplary value is 10.

[0095] For non-abnormal monitoring points (ordinary monitoring points), set the filter weight to 1+ This weight is positively correlated with the attention factor: when the attention factor When the risk level is high (overall structural risk is high and showing a deteriorating trend), the weight value increases, the smoothing effect weakens, more data details are retained, and potential structural anomaly signals are avoided; when the focus factor is high... When the risk is low (the overall structural risk is small and stable), the weight value decreases, the smoothing effect is enhanced, and the invalid fluctuations caused by environmental noise and sensor interference are effectively filtered out.

[0096] Based on the aforementioned differentiated filtering weights, a weighted moving average filtering method using historical data with a window width of 5 time points is employed to filter the stress time series data of all monitoring points, resulting in filtered stress time series data. Within the filtering window, the weights of all historical data points except the current monitoring time are fixed at 1, and the final filtering result is the weighted arithmetic mean of all data points within the window.

[0097] Based on the above technical solution, this embodiment identifies risk areas by spatial clustering of abnormal monitoring points, thereby achieving a quantitative characterization of the spatial distribution characteristics of structural anomalies. It calculates the overall structural risk index by comprehensively considering the number, scale, and spatial distribution density of risk areas, and determines the risk evolution trend by combining the temporal variation characteristics of the risk area, thus achieving dynamic assessment and evolution prediction of the overall structural risk status. Based on the overall structural risk index and risk evolution trend, it determines the factors of concern and sets differentiated filtering weights accordingly. This achieves low-intensity filtering (preserving details) of monitoring data at high-risk times and in high-risk areas, and high-intensity filtering (suppressing noise) of low-risk data, significantly improving the adaptability of filtering and data quality. It effectively balances the contradiction between noise suppression and anomaly preservation, ensuring the integrity and reliability of key structural anomaly information.

[0098] In one possible implementation, the above steps perform feature fusion on the filtered stress time series data, displacement time series data, engineering project design data, construction data, and environmental monitoring data to obtain feature fused data, including: Step 601: Spatiotemporally align the filtered stress time series data with the displacement time series data, design data, construction data, and environmental monitoring data.

[0099] Specifically, the filtered stress time series data, displacement time series data, construction process data, and environmental monitoring data from all monitoring points are unified to the same timestamp reference. This can be achieved through nearest neighbor matching, linear interpolation, or timestamp-based resampling methods, ensuring the comparability of multi-source data at the same time.

[0100] For displacement data, environmental monitoring data, etc., their spatial coordinates can be mapped to the vicinity of the corresponding stress monitoring point. Spatial interpolation (such as inverse distance weighted interpolation, Kriging interpolation) or nearest neighbor algorithm can be used to ensure that the comparison is of the stress and deformation and environmental conditions of the same physical location or neighboring area.

[0101] Step 602: For the spatiotemporally aligned data: Extract stress statistical features from the filtered stress time series data. Extract design benchmark features from the design data, and determine design deviation features based on the differences between the filtered stress time series data and the design benchmark features. Extract construction stage identification features and construction load application features from the construction data. Extract environmental statistical features from the environmental monitoring data. Extract displacement statistical features from the displacement time series data. Stress statistical features include mean, standard deviation, peak value, frequency, and fluctuation range. Displacement statistical features include displacement change, maximum displacement, and displacement rate features. Environmental statistical features include temperature influence features, humidity influence features, wind speed influence features, vibration influence features, and rainfall influence features.

[0102] Among them, the statistical characteristics of stress include statistical quantities such as mean, standard deviation, peak value, frequency and fluctuation range, which are used to characterize the static distribution and dynamic change characteristics of structural stress. They can be obtained by calculating the statistical parameters of filtered stress data within a preset time window (such as the past 24 hours).

[0103] Displacement statistical characteristics include displacement change, maximum displacement, and displacement rate characteristics, which are used to characterize the deformation response of a structure. They can be obtained by calculating the change in current displacement relative to the initial displacement or the displacement at the previous moment, the maximum displacement value within the monitoring period, and the displacement change rate.

[0104] Design baseline features include design stress indices, allowable stress thresholds, material properties, and structural geometric parameters. By comparing the differences between the filtered stress time series data and the design baseline features (such as the ratio of actual stress to allowable stress, and the difference between actual strain and theoretically calculated strain), design deviation features are calculated to characterize the degree of deviation between the actual stress state and the design expectation.

[0105] Construction stage identification features are used to identify the current construction stage (such as foundation construction stage, main structure construction stage, decoration stage, etc.), which can be represented by unique thermal coding or stage number; construction load application features are used to characterize the type (such as pouring, tensioning, hoisting) and magnitude of the construction load applied at the current moment, which can be obtained by extracting from construction logs or monitoring by sensors.

[0106] Environmental statistical characteristics include the influence of temperature, humidity, wind speed, vibration, and rainfall. These characteristics can be represented by parameters such as the mean, extreme values, and rate of change of temperature, humidity, wind speed, vibration acceleration, and rainfall within the statistical window. Furthermore, the correlation coefficient between environmental factors and stress response can be calculated as a feature.

[0107] Step 603: Based on dimensionality reduction technology, the extracted stress statistical features, displacement statistical features, design deviation features, construction stage identification features, construction load application features, temperature influence features, humidity influence features, and wind speed influence features are fused into a low-dimensional feature vector.

[0108] Dimensionality reduction techniques such as Principal Component Analysis (PCA), Linear Discriminant Analysis (LDA), t-distributed stochastic neighbor embedding (t-SNE), or autoencoders are used to map the extracted high-dimensional feature set (which may contain tens to hundreds of features) to a low-dimensional space (e.g., 10-20 dimensions), forming low-dimensional feature vectors. This process eliminates redundancy and correlation between features while preserving key structural information, reducing the computational complexity of subsequent data processing and storage. During dimensionality reduction, the optimal number of dimensions can be determined based on the contribution rate of features to variance or reconstruction error, ensuring that the fused feature vectors retain more than 95% of the information from the original data.

[0109] Step 604: Use the low-dimensional feature vector as feature fusion data.

[0110] Optionally, this low-dimensional feature vector serves as the basic data unit for establishing subsequent data traceability links. It contains not only the core information of the structural stress state but also integrates multi-dimensional contextual information such as design, construction, and environment, thus achieving a comprehensive representation of the project status.

[0111] Based on the above technical solutions, this embodiment ensures the comparability and consistency of heterogeneous multi-source data through spatiotemporal alignment, achieves a comprehensive characterization of structural stress state, design deviation, construction activities and environmental conditions through multidimensional feature extraction, and realizes the effective fusion and information compression of high-dimensional features through dimensionality reduction technology. While retaining key structural information, it significantly reduces data complexity, providing a standardized, low-dimensional and information-rich data foundation for the subsequent establishment of a full-cycle data traceability link.

[0112] In one possible implementation, the above steps, based on feature fusion data, establish a data traceability link to characterize the evolution of the structural stress state throughout the entire lifecycle of an engineering project, including: Step 701: Link and store the feature fusion data with the design phase data, construction phase data, acceptance phase data and operation and maintenance phase data at the corresponding monitoring time to build a full-cycle database.

[0113] Specifically, the aforementioned low-dimensional feature fusion data is linked and stored with stage data such as design drawings, design calculation sheets, construction logs, quality acceptance reports, and operation and maintenance inspection records at the corresponding monitoring time, to establish a unified database covering the entire life cycle of engineering project design, construction, acceptance, and operation and maintenance.

[0114] Optionally, the database can be a relational database (such as MySQL or PostgreSQL) or a time-series database (such as InfluxDB or TimescaleDB), supporting efficient storage and querying of massive amounts of monitoring data. Data association can be achieved through foreign keys or timestamp-spatial location composite indexes, ensuring the logical association between feature fusion data and business data at each stage.

[0115] Step 702: Add timestamp labels, spatial location labels, and process stage labels to the feature fusion data and data at each stage to form a multi-dimensional data index.

[0116] Add timestamp tags (accurate to seconds, minutes, or hours, depending on the data sampling frequency), spatial location tags (3D coordinates of monitoring points or regional codes), and process stage tags (design / construction / acceptance / operation and maintenance stage identifiers, which can be further subdivided into specific processes such as foundation pouring, prestressing tensioning, etc.) to each record in the database to build a data index structure that supports multi-dimensional cross-retrieval of time, space, and process.

[0117] For example, timestamp labels can use Unix timestamps or ISO8601 format, spatial location labels can use WGS84 coordinate system or engineering local coordinate system, and process stage labels can use a standardized coding system.

[0118] Step 703: Establish a cross-stage data traceability path based on the multi-dimensional data index.

[0119] Optionally, based on multi-dimensional data indexing, a multi-level traceability path can be established, from design parameters to construction activities, from construction activities to structural response, and from structural response to environmental impact.

[0120] For example, the structural stress response state at the time of a specific construction activity (such as a concrete pour) can be traced based on timestamp tags; the stress evolution of a specific area (such as a bridge span) during the design, construction, and operation and maintenance stages can be traced based on spatial location tags; and the structural state changes at different construction stages (such as before pouring vs. after pouring) can be compared based on process stage tags. The traceability path can be established using graph databases (such as Neo4j) or linked list structures, supporting forward tracing (from cause to effect) and reverse tracing (from effect to cause).

[0121] Step 704: Generate a visual traceability map based on the cross-stage data traceability path. The visual traceability map is used to show the relationship between the structural stress state and design benchmarks, construction activities, and environmental factors at different stages.

[0122] Based on cross-stage data tracing paths, a visual tracing map is generated using graph visualization technologies (such as force-directed graphs, Sankey diagrams, or timeline charts). This map uses a timeline as the main line and spatial locations as nodes to show the correlation between the structural stress state and design benchmarks, construction activities, and environmental factors at different stages. For example, color coding can be used to represent stress magnitude (e.g., blue for low stress, red for high stress), node size to represent risk level, and lines to represent causal relationships or time sequence. Users can perform drill-down (from project level to component level), tracing (from abnormal responses to construction activities), and comparative analysis (comparing states in different regions or periods) through an interactive interface (such as a web or mobile application), intuitively presenting the evolution of the structural stress state throughout the entire project lifecycle.

[0123] Based on the above technical solutions, this embodiment realizes centralized management and associated storage of multi-source heterogeneous data by constructing a full-cycle database, establishes a three-dimensional retrieval system of time-space-process through multi-dimensional data indexing, realizes the causal relationship of the entire chain from design, construction to operation and maintenance through cross-stage data traceability paths, and provides intuitive and interactive data display methods through visual traceability maps. Thus, it realizes accurate traceability and visualization of the evolution of structural stress state throughout the entire life cycle of the engineering project, and provides comprehensive and systematic data support for engineering structure safety assessment, accident tracing and maintenance decision-making.

[0124] Please see Figure 2 This document illustrates a system architecture diagram of a project data full-cycle traceability system based on multi-source data fusion, according to an embodiment of the present invention. The system includes: a data acquisition unit 201, an anomaly identification unit 202, a filtering unit 203, a feature fusion unit 204, and a link establishment unit 205. The units communicate bidirectionally via communication links, ensuring real-time interaction between the acquired data and analysis results. The communication links can employ wired or wireless transmission methods to meet the communication needs of different monitoring scenarios.

[0125] The data acquisition unit 201 is used to acquire stress time series data and displacement time series data of key stress-bearing parts of the engineering structure; Anomaly identification unit 202 is used to identify abnormal monitoring points at each monitoring time based on the degree of deviation of the temporal characteristics of each monitoring point and the abnormal synchronization correlation characteristics between the monitoring points and other monitoring points; The filtering unit 203 is used to adaptively filter the stress time series data of each monitoring point according to the spatial distribution concentration of the abnormal monitoring points and the overall risk evolution trend of the structure, so as to obtain the filtered stress time series data; the overall risk evolution trend of the structure is used to characterize the degree of expansion or contraction of the abnormal risk area of ​​the structure over time. The feature fusion unit 204 is used to perform feature fusion on the filtered stress time series data, the displacement time series data, the design data of the engineering project, the construction data and the environmental monitoring data to obtain feature fused data. Link establishment unit 205 is used to establish a data traceability link based on the feature fusion data to characterize the evolution of the structural stress state throughout the entire life cycle of an engineering project.

[0126] It should be noted that the order of the embodiments described above is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. The processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0127] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

Claims

1. A method for full-lifecycle traceability of project data based on multi-source data fusion, characterized in that, include: Collect stress time series data and displacement time series data of key stress-bearing parts of the engineering structure; Based on the degree of deviation of the temporal characteristics of each monitoring point and the abnormal synchronization correlation characteristics between the monitoring point and other monitoring points, abnormal monitoring points are identified at each monitoring time. Based on the spatial distribution concentration of abnormal monitoring points and the overall risk evolution trend of the structure, adaptive filtering is performed on the stress time series data of each monitoring point to obtain filtered stress time series data. The overall structural risk evolution trend is used to characterize the degree of expansion or contraction of structural anomaly risk regions over time. The filtered stress time series data is fused with the displacement time series data, the design data, construction data and environmental monitoring data of the project to obtain feature fused data; Based on the aforementioned feature fusion data, a data traceability link is established to characterize the evolution of the structural stress state throughout the entire life cycle of an engineering project.

2. The project data full-cycle traceability method based on multi-source data fusion according to claim 1, characterized in that, Based on the degree of deviation of the temporal characteristics of each monitoring point and the abnormal synchronization correlation characteristics between the monitoring points and other monitoring points, abnormal monitoring points at each monitoring time are identified, including: For each monitoring time, based on the degree of deviation of the temporal characteristics of each monitoring point at that monitoring time, monitoring points that meet the preset anomaly conditions are selected from all monitoring points and determined as candidate points for temporal anomalies; the degree of deviation of the temporal characteristics is used to characterize the deviation of the monitoring point from historical trend data at the current monitoring time. Based on the abnormal synchronization correlation characteristics between the time-series anomaly candidate points and other monitoring points, abnormal monitoring points are determined from the time-series anomaly candidate points; the abnormal synchronization correlation characteristics are used to characterize the degree of coordination between the time-series anomaly candidate points and surrounding monitoring points in the same time period when stress abrupt changes occur.

3. The project data full-cycle traceability method based on multi-source data fusion according to claim 2, characterized in that, The process of determining the degree of deviation of the time series features includes: Acquire historical stress time-series data of the monitoring point prior to the monitoring time; The historical stress time series data is processed based on the trend extraction method to obtain trend change data; Calculate the residual data of the stress data at the current monitoring moment relative to the trend change data; The degree of deviation of the time series features is determined based on the distribution and trend characteristics of the residual data.

4. The project data full-cycle traceability method based on multi-source data fusion according to claim 2, characterized in that, The process of determining the abnormal synchronization association characteristics includes: Obtain the sequence of deviations in temporal characteristics between the candidate temporal anomalies and the other monitoring points within the same time period; Analyze the changing trend of the deviation degree sequence of the time series features, and determine the synchronous change parameters based on the similarity of the changing trends; The abnormal synchronization correlation characteristics are determined based on the distance parameters between the candidate time-series anomalies and the other monitoring points, as well as the synchronization change parameters.

5. The project data full-cycle traceability method based on multi-source data fusion according to claim 1, characterized in that, Based on the spatial distribution concentration of abnormal monitoring points and the overall risk evolution trend of the structure, adaptive filtering is performed on the stress time series data of each monitoring point, including: Based on spatial distribution characteristics, abnormal monitoring points are clustered to obtain risk regions; the risk regions are used to characterize a set of monitoring points that are spatially continuous and have synchronous stress anomalies. The overall structural risk index is determined based on the number of risk areas, the number of monitoring points within each risk area, and the spatial distribution density between risk areas. The overall risk evolution trend of the structure is determined based on the temporal variation characteristics of the risk area; the risk area is the total number of abnormal monitoring points contained in all risk areas at each monitoring time. Based on the overall structural risk index and the overall structural risk evolution trend, focus factors are determined; Differentiated filtering weights are set for different monitoring points based on the aforementioned factors of interest, and adaptive filtering is performed based on these differentiated filtering weights.

6. The project data full-lifecycle traceability method based on multi-source data fusion according to claim 5, characterized in that, The overall structural risk index is determined based on the number of risk areas, the number of monitoring points within each risk area, and the spatial distribution density between risk areas, including: Determine the total number of risk areas, and count the number of monitoring points within each risk area; The average distance between risk areas is determined based on the mean of the distances between all risk areas, and the spatial distribution density parameter is determined based on the average distance; the spatial distribution density parameter is used to characterize the degree of spatial concentration of risk areas. For each risk area, the risk scale parameter of the risk area is determined based on the number of monitoring points in the risk area and the average stress anomaly degree of all monitoring points in the risk area. The risk scale parameters of all risk areas are summed to obtain the overall risk area scale parameter; The overall structural risk index is determined based on the overall scale parameter of the risk area and the spatial distribution density parameter.

7. The project data full-lifecycle traceability method based on multi-source data fusion according to claim 5, characterized in that, Based on the overall structural risk index and the overall structural risk evolution trend, the factors of concern are determined, including: Obtain the area of ​​the risk zone at each monitoring time; the area of ​​the risk zone is the total number of abnormal monitoring points contained in all risk zones at the corresponding monitoring time. The change in risk area is determined by the difference between the risk area at the previous monitoring time and the risk area at the current monitoring time. The trend parameters of risk area change are determined based on the slope of the trend line obtained by fitting the risk area area at each monitoring time. The overall risk evolution trend of the structure is determined by multiplying the change in the area of ​​the risk region by the trend parameter of the change in the area of ​​the risk region. The product of the overall structural risk index and the overall structural risk evolution trend is normalized to obtain the attention factor.

8. The project data full-lifecycle traceability method based on multi-source data fusion according to claim 1, characterized in that, The filtered stress time series data is fused with the displacement time series data, the project design data, construction data, and environmental monitoring data to obtain feature-fused data, including: The filtered stress time series data is spatiotemporally aligned with the displacement time series data, the design data, the construction data, and the environmental monitoring data. For the spatiotemporally aligned data: stress statistical features are extracted from the filtered stress time series data; design benchmark features are extracted from the design data, and design deviation features are determined based on the difference between the filtered stress time series data and the design benchmark features; construction stage identification features and construction load application features are extracted from the construction data; environmental statistical features are extracted from the environmental monitoring data; displacement statistical features are extracted from the displacement time series data; the stress statistical features include mean, standard deviation, peak value, frequency, and fluctuation range; the displacement statistical features include displacement change, maximum displacement, and displacement rate features; the environmental statistical features include temperature influence features, humidity influence features, wind speed influence features, vibration influence features, and rainfall influence features. Based on dimensionality reduction technology, the extracted stress statistical features, displacement statistical features, design deviation features, construction stage identification features, construction load application features, temperature influence features, humidity influence features, and wind speed influence features are fused into a low-dimensional feature vector. The low-dimensional feature vector is used as the feature fusion data.

9. The project data full-lifecycle traceability method based on multi-source data fusion according to claim 8, characterized in that, Based on the aforementioned feature fusion data, a data traceability link is established to characterize the evolution of the structural stress state throughout the entire lifecycle of an engineering project, including: The feature fusion data is associated and stored with the design phase data, construction phase data, acceptance phase data and operation and maintenance phase data at the corresponding monitoring time to construct a full-cycle database. Add timestamp tags, spatial location tags, and process stage tags to the feature fusion data and data at each stage to form a multi-dimensional data index; Based on the multi-dimensional data index, establish a cross-stage data tracing path; Based on the cross-stage data tracing path, a visual tracing map is generated; the visual tracing map is used to show the correlation between the structural stress state and design benchmarks, construction activities and environmental factors at different stages.

10. A project data full-lifecycle traceability system based on multi-source data fusion, characterized in that, include: The data acquisition unit is used to collect stress time-series data and displacement time-series data of key stress-bearing parts of the engineering structure; An anomaly identification unit is used to identify abnormal monitoring points at each monitoring time based on the degree of deviation of the temporal characteristics of each monitoring point and the abnormal synchronization correlation characteristics between the monitoring point and other monitoring points. The filtering unit is used to adaptively filter the stress time series data of each monitoring point according to the spatial distribution concentration of abnormal monitoring points and the overall risk evolution trend of the structure, so as to obtain the filtered stress time series data. The overall structural risk evolution trend is used to characterize the degree of expansion or contraction of structural anomaly risk regions over time. The feature fusion unit is used to perform feature fusion on the filtered stress time series data, the displacement time series data, the design data of the engineering project, the construction data, and the environmental monitoring data to obtain feature fused data. The link establishment unit is used to establish a data traceability link based on the feature fusion data to characterize the evolution of the structural stress state throughout the entire life cycle of an engineering project.

Images