Fatigue damage detection method and system for large grid structure

Abstract

The invention discloses a fatigue damage detection method and system for a large-scale grid structure, and relates to the related field of structure health monitoring, and the method comprises the steps: arranging a plurality of types of sensors based on the large-scale grid structure to carry out synchronous response collection, and obtaining a structure response data set; synchronously obtaining real-time load data and combining the real-time load data with the structure parameters to construct a grid structure digital twinborn model for simulation calculation, and obtaining a theoretical structure response data set; comparing the theoretical structure response data set with the structure response data set, performing exception identification according to a comparison result, and identifying and positioning a response exception region; evolution analysis is carried out based on the real-time load data, time sequence change characteristics are constructed, fatigue damage evolution is carried out based on the response abnormal area in combination with the time sequence change characteristics, and a fatigue damage detection report is generated. The technical problem that the accuracy of damage positioning and evaluation is insufficient in existing large grid structure fatigue damage detection is solved, and the technical effect of improving the accuracy of damage positioning and evaluation is achieved.

Description

Technical Field

[0001] This application relates to the field of structural health monitoring, and in particular to a method and system for detecting fatigue damage in large space frame structures. Background Technology

[0002] Large-scale space frame structures are widely used in large public buildings such as stadiums and airport terminals. Their fatigue damage status directly affects the safe operation of the structure and the safety of people and property. Therefore, accurate fatigue damage detection is of paramount importance. Currently, the mainstream technology for fatigue damage detection in large-scale space frame structures mainly relies on traditional sensor monitoring combined with empirical formulas for assessment. Specifically, this involves deploying single-type sensors at key structural locations to collect response data, and then using empirical formulas such as material fatigue curves to estimate the degree of damage. However, existing methods do not adequately consider the complex stress characteristics of the structure and the coupling effect of dynamic loads, resulting in a lack of spatiotemporal synchronization in the detection data, making it difficult to accurately locate the damaged area, and leading to a significant deviation between the fatigue damage assessment results and the actual structural condition.

[0003] Currently, the fatigue damage detection of large space frame structures suffers from insufficient accuracy in damage localization and assessment. Summary of the Invention

[0004] This application provides a fatigue damage detection method and system for large-scale space frame structures. It employs multiple sensors deployed on the large space frame structure to synchronously collect response data, obtaining a structural response dataset. Real-time load data is simultaneously acquired and combined with structural parameters to construct a digital twin model. A theoretical structural response dataset is obtained through simulation calculation. The two datasets are compared and anomalies are identified, abnormal response areas are located, evolutionary analysis of real-time load data is performed to construct temporal variation characteristics, and fatigue damage evolution analysis is conducted based on the abnormal response areas to generate a fatigue damage detection report. These techniques solve the technical problem of insufficient accuracy in damage location and assessment in existing fatigue damage detection methods for large-scale space frame structures, achieving the technical effect of improving the accuracy of damage location and assessment.

[0005] This application provides a fatigue damage detection method for large-scale space frame structures, comprising: deploying multiple types of sensors on the large-scale space frame structure for synchronous response acquisition to obtain a structural response dataset; synchronously acquiring real-time load data of the large-scale space frame structure and constructing a digital twin model of the space frame structure based on the structural parameters of the large-scale space frame structure; performing simulation calculations through the digital twin model of the space frame structure to obtain a theoretical structural response dataset; comparing the theoretical structural response dataset with the structural response dataset, identifying anomalies based on the comparison results, and locating abnormal response areas; performing evolution analysis based on the real-time load data to construct time-series variation characteristics; performing fatigue damage evolution analysis on the large-scale space frame structure based on the abnormal response areas and the time-series variation characteristics, and generating a fatigue damage detection report.

[0006] In a possible implementation, real-time load data of a large-scale space frame structure is acquired synchronously and combined with the structural parameters of the large-scale space frame structure to construct a digital twin model of the space frame structure. Simulation calculations are then performed using the digital twin model to obtain a theoretical structural response dataset. The following processes are then executed: Real-time monitoring data streams of the large-scale space frame structure are synchronously accessed and used as dynamic external load inputs; these dynamic external load inputs are mapped to an initial structural baseline model, and a structure-load interaction analysis is performed based on the initial structural baseline model to generate dynamic structure-load interaction state parameters; a digital twin model of the space frame structure is constructed based on these dynamic structure-load interaction state parameters and the structural parameters; the digital twin model of the space frame structure is driven to perform real-time simulation cycles at multiple time steps to obtain a mechanical equilibrium dataset; multi-dimensional feature analysis is performed based on the mechanical equilibrium dataset to determine multi-dimensional simulation feature values; and the multi-dimensional simulation feature values ​​are aligned according to the structural parameters to construct a theoretical structural response dataset.

[0007] In a possible implementation, the construction process of the initial structural baseline model involves the following steps: retrieving the engineering archive of large-scale space frame structures, traversing the archive for geometric analysis, and constructing geometric topological relationships; performing a three-dimensional spatial scan based on the geometric topological relationships to construct geometric shape spatial coordinates; fusing and comparing the structural parameters with the geometric shape spatial coordinates, extracting difference parameters based on the comparison results, correcting the geometric shape spatial coordinates according to the difference parameters, and generating three-dimensional solid space beam and rod data; and performing mesh generation and assignment based on the three-dimensional solid space beam and rod data to construct the initial structural baseline model.

[0008] In a possible implementation, the theoretical structural response dataset is compared with the structural response dataset. Anomalies are identified and located based on the comparison results. The following processing steps are performed: The theoretical structural response dataset and the structural response dataset are timestamped to generate first aligned data; the theoretical structural response dataset and the structural response dataset are aligned in three-dimensional space to generate second aligned data; the first aligned data and the second aligned data are fused to construct a time-space analysis framework; the theoretical structural response dataset and the structural response dataset are traversed point-by-point according to the time-space analysis framework to calculate differences and generate a residual data field; multidimensional anomaly feature analysis is performed based on the residual data field to determine multidimensional anomaly features; multiple anomaly data points are identified according to the multidimensional anomaly features; spatiotemporal analysis is performed on the multiple anomaly data points, and multiple spatiotemporally adjacent anomaly data points are extracted for spatial clustering to delineate the response anomaly region.

[0009] In a possible implementation, a time-space analysis framework is constructed by fusing the first alignment data and the second alignment data, and the following processes are performed: a multidimensional data index structure is created using timestamps and three-dimensional spatial coordinates as a joint primary key; a spatial mapping relationship is defined based on the second alignment data, and the first alignment data is filled into the multidimensional data index structure according to the spatial mapping relationship; time slicing is performed based on the multidimensional data index structure to determine multiple time slices, and spatial queries are performed according to the multiple time slices to determine spatial distribution data; the time-space analysis framework is constructed by combining the multiple time slices and the spatial distribution data into an array.

[0010] In a possible implementation, the theoretical structural response dataset and the structural response dataset are traversed point-by-point according to the time-space analysis framework to calculate the difference, generating a residual data field. The following processing is then performed: Based on the time-space analysis framework, the theoretical structural response dataset and the structural response dataset are analyzed according to a preset time period to extract the measured response time history curve and the theoretical response time history curve; the theoretical response time history curve is used as a benchmark to perform time-by-time subtraction on the measured response time history curve to generate a residual time series; the residual time series is traversed to perform measurement point source space analysis to determine the measurement point spatial coordinates; the residual time series is traversed to perform measurement point source time analysis to determine the measurement point time parameters; and the residual data field is constructed based on the measurement point spatial coordinates and the measurement point time parameters according to the spatiotemporal dimension.

[0011] In a possible implementation, multidimensional anomaly feature analysis is performed based on the residual data field to determine multidimensional anomaly features. Multiple anomalous data points are identified based on these multidimensional anomaly features, and the following processing is performed: A dynamic amplitude threshold is set to traverse the residual data field for time-domain amplitude analysis to determine time-domain anomaly features; point comparison is performed based on the time-domain anomaly features to determine time-domain amplitude anomalies; a frequency band energy threshold is set to traverse the residual data field for frequency-domain energy analysis to determine frequency-domain energy anomaly features; point comparison is performed based on the frequency-band energy anomaly features to determine frequency-domain energy anomalies; statistical analysis is performed by traversing the residual data field according to a moving time window to determine probability distribution features; point comparison is performed based on the probability distribution features to determine statistical distribution anomalies; and the time-domain amplitude anomalies, frequency-domain energy anomalies, and statistical distribution anomalies are integrated to obtain the multiple anomalous data points.

[0012] In a possible implementation, evolution analysis is performed based on the real-time load data to construct time-series change characteristics, and the following processing is performed: polymorphic separation and identification is performed based on the real-time load data to obtain static load components and dynamic load fluctuation components; the dynamic load fluctuation components are cyclically counted according to the load cycle to generate the load cycle count and load cycle occurrence frequency; amplitude analysis is performed on the dynamic load fluctuation components to divide them into multiple amplitude levels, and the load cycle count and load cycle occurrence frequency are analyzed according to the multiple amplitude levels to construct a load amplitude-frequency distribution spectrum; statistics are performed based on the load amplitude-frequency distribution spectrum to extract key load characteristic indicators; and changes in the key load characteristic indicators are tracked in chronological order to construct the time-series change characteristics.

[0013] In a possible implementation, fatigue damage evolution of a large-scale space frame structure is performed based on the abnormal response region and the temporal variation characteristics to generate a fatigue damage detection report. The following processing is then performed: long-term field measurements are conducted based on the abnormal response region to extract an abnormal response dataset; the abnormal response dataset is correlated with the temporal variation characteristics to generate correlation factors; local response analysis is performed on the abnormal response region according to the correlation factors to generate local response patterns; fatigue damage is extrapolated based on the local response patterns to construct a local stress spectrum; fatigue accumulation analysis is performed based on the local stress spectrum to calculate the cumulative fatigue degree and perform damage evolution to generate remaining fatigue life data; the remaining fatigue life data is spatially distributed and mapped according to the abnormal response region to construct the fatigue damage detection report.

[0014] This application also provides a fatigue damage detection system for large-scale space frame structures, comprising: a structural response data acquisition module for synchronously acquiring responses from multiple types of sensors deployed on the large-scale space frame structure to obtain a structural response dataset; a space frame structure digital twin model construction module for synchronously acquiring real-time load data of the large-scale space frame structure and constructing a digital twin model of the space frame structure based on the structural parameters of the large-scale space frame structure, and performing simulation calculations through the space frame structure digital twin model to obtain a theoretical structural response dataset; an anomaly comparison module for comparing the theoretical structural response dataset with the structural response dataset, identifying anomalies based on the comparison results, and locating abnormal response areas; and a fatigue damage evolution module for performing evolution analysis based on the real-time load data, constructing time-series change characteristics, performing fatigue damage evolution analysis on the large-scale space frame structure based on the abnormal response areas and the time-series change characteristics, and generating a fatigue damage detection report.

[0015] The proposed fatigue damage detection method and system for large-scale space frame structures involves several steps. First, multiple sensors are deployed on the large-scale space frame structure to synchronously acquire responses, obtaining a structural response dataset. Next, real-time load data of the large-scale space frame structure is simultaneously acquired and combined with its structural parameters to construct a digital twin model of the space frame structure. Simulation calculations are then performed using this digital twin model to obtain a theoretical structural response dataset. The theoretical structural response dataset is then compared with the proposed structural response dataset, and anomalies are identified and located based on the comparison results. Finally, evolution analysis is performed on the real-time load data to construct temporal variation characteristics. Based on the identified response anomalies and these temporal variation characteristics, the fatigue damage evolution of the large-scale space frame structure is analyzed, generating a fatigue damage detection report. Through this process, the proposed method and system achieve the technical effect of improving the accuracy of damage location and assessment. Attached Figure Description

[0016] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings of the embodiments of the present invention will be briefly described below. Flowcharts are used in this application to illustrate the operations performed by the system according to the embodiments of the present application. It should be understood that the preceding or following operations are not necessarily performed precisely in sequence. Instead, various steps can be processed in reverse order or simultaneously as needed. Furthermore, other operations can be added to these processes, or one or more steps can be removed from these processes.

[0017] Figure 1 This is a flowchart illustrating a fatigue damage detection method for large space frame structures provided in an embodiment of this application.

[0018] Figure 2 This is a structural schematic diagram of a fatigue damage detection system for large space frame structures provided in an embodiment of this application.

[0019] Figure labeling: 10 Structural response data acquisition module, 20 Digital twin model construction module for space frame structure, 30 Anomaly comparison module, 40 Fatigue damage evolution module. Detailed Implementation

[0020] To further illustrate the technical means and effects of the present invention in achieving its intended purpose, the following detailed description of the specific implementation methods, structures, features, and effects of the present invention, in conjunction with the accompanying drawings and preferred embodiments, is provided below.

[0021] This application provides a fatigue damage detection method for large space frame structures, such as... Figure 1 As shown, the method includes: Step S100: Based on the large-scale grid structure, multiple types of sensors are deployed to collect synchronous responses and obtain a structural response dataset.

[0022] Specifically, by deploying various types of sensors on a large space frame structure, the response signals of the structure under load are simultaneously collected according to a unified time reference, forming a structured structural response dataset. This dataset is a structured collection of data containing fields such as sensor ID, acquisition timestamp, installation location coordinates, and response parameter values. The large space frame structure refers to a spatial load-bearing structure formed by multiple beam-and-rod members connected by nodes, such as the roof space frame of a stadium. The various types of sensors are devices used to collect different structural response parameters, such as strain sensors, acceleration sensors, displacement sensors, and tilt sensors.

[0023] Step S200: Simultaneously acquire real-time load data of the large-scale space frame structure and construct a digital twin model of the space frame structure based on its structural parameters. Perform simulation calculations using the digital twin model of the space frame structure to obtain a theoretical structural response dataset.

[0024] Specifically, this involves acquiring real-time load data on large-scale space frame structures, i.e., external force data experienced by the structure during operation, such as the structure's self-weight, the weight of fixed equipment, wind loads, vibration loads, and temporary loads. This data is then combined with the structure's inherent parameters—parameters characterizing the inherent properties of the space frame structure, including member length, cross-sectional dimensions, node coordinates, elastic modulus, Poisson's ratio, yield strength, density, node stiffness, and member bending stiffness—to construct a digital twin model of the space frame structure that reflects its true state. This digital twin model is a virtual model of the space frame structure, built using digital technology, that maps and behaves identically to the physical structure in real time, simulating the mechanical response of the structure under different load conditions. Simulation calculations are performed using this model to obtain theoretical response data corresponding to the measured data, forming a theoretical structural response dataset. The data format and fields of this dataset are consistent with the structural response dataset and are used for comparison with the measured data.

[0025] In one possible implementation, real-time load data of a large-scale space frame structure is synchronously acquired and combined with the structural parameters of the large-scale space frame structure to construct a digital twin model of the space frame structure. Simulation calculations are then performed using the digital twin model to obtain a theoretical structural response dataset. Step S200 further includes step S210, synchronously accessing the real-time monitoring data stream of the large-scale space frame structure and using this real-time monitoring data stream as a dynamic external load input. Specifically, a real-time data acquisition gateway is used as the core device, accessing the real-time data streams of various load monitoring devices via wired or wireless communication protocols. The input data is timestamped using a time synchronization protocol to ensure consistency between the time base of the load data and the structural response data. The acquired load data is preprocessed, including data filtering, outlier removal, and missing value completion, ultimately forming standardized dynamic external load input data.

[0026] Step S220: The dynamic external load input is mapped to the initial structural reference model. Based on the initial structural reference model, a structure-load interaction analysis is performed to generate dynamic structure-load interaction state parameters. Specifically, the preprocessed dynamic external load input data is mapped to the corresponding stress locations in the initial structural reference model according to the geometric topology of the structure. The finite element method is used to perform a structure-load interaction analysis on the model, calculating the stress, strain, displacement, and other mechanical parameters of the structure under the load. Based on these mechanical parameters, key parameters characterizing the interaction state between the structure and the load are extracted to form dynamic structure-load interaction state parameters.

[0027] Step S230: Construct a digital twin model of the space frame structure based on the dynamic structure-load interaction state parameters and the structural parameters. Specifically, based on the initial structural baseline model, the dynamic structure-load interaction state parameters and structural parameters are fused. A combination of data-driven and physical modeling methods is used to correct the model's boundary conditions, material parameters, node stiffness, and other parameters, ensuring that the model's mechanical properties match the actual structural state. A real-time mapping relationship between the model and the physical structure is established to ensure that the model can update its state synchronously when dynamic external loads change.

[0028] Step S240: Drive the digital twin model of the space frame structure to perform real-time simulation loops according to multiple time steps to obtain a mechanical equilibrium dataset. Specifically, an explicit time integration algorithm is used to drive the digital twin model of the space frame structure for real-time simulation. A fixed time step is set. Within each time step, the dynamic external load at that moment is input, and the mechanical equilibrium equations of each component of the structure are calculated to obtain the mechanical equilibrium data for each time step. An iterative solution method, such as the Newton-Raphson method, is used during the simulation to ensure that the calculation results of each time step converge.

[0029] Step S250: Perform multi-dimensional feature analysis based on the mechanical equilibrium dataset to determine multi-dimensional simulation feature values. Specifically, multi-dimensional feature extraction is performed on the mechanical equilibrium dataset using methods such as statistical analysis, time-domain analysis, and frequency-domain analysis to extract key feature values ​​that characterize the structural response. Time-domain features include peak value, valley value, mean, variance, root mean square, peak factor, kurtosis, and skewness, while frequency-domain features include natural frequency, mode shape, band energy, and spectral peak value. Redundant features are eliminated through correlation analysis or principal component analysis, retaining multi-dimensional simulation feature values ​​sensitive to structural damage.

[0030] Step S260: Align the multidimensional simulation feature values ​​according to the structural parameters to construct a theoretical structural response dataset. Specifically, align the multidimensional simulation feature values ​​with the physical location of the structure according to the geometric parameters in the structural parameters, such as node coordinates and component numbers. Construct a theoretical structural response dataset with consistent fields and a unified structure, referring to the format of the structural response dataset. Ensure that the dimensions, indexes, and physical meanings of the theoretical data and the measured data are completely matched to facilitate comparative analysis.

[0031] In one possible implementation, the initial structural baseline model construction process, step S220, further includes step S221: retrieving the engineering archive database of the large-scale space frame structure, traversing the archive database for geometric analysis, and constructing geometric topological relationships. Specifically, the engineering archive database, including design drawings, construction records, and as-built data of the large-scale space frame structure, is retrieved through the engineering archive management system. Drawing parsing algorithms, such as OpenCV-based image recognition and CAD file secondary development interfaces, are used to extract geometric elements such as components, nodes, and supports from the drawings. By traversing all geometric elements, the connection relationships between components and nodes, and between nodes and supports, are established, such as the node numbers connecting the two ends of a component and the correspondence between nodes and supports, forming a structured geometric topological relationship table.

[0032] Step S222: Perform a three-dimensional spatial scan based on the aforementioned geometric topology to construct the geometric shape spatial coordinates. Specifically, based on the geometric topology table, use three-dimensional laser scanning technology to perform a full-space scan of the physical grid structure to obtain point cloud data of the structural surface. Use a point cloud registration algorithm to stitch and fuse the point cloud data from multiple perspectives, removing noise points. According to the geometric topology, match the point cloud data with the geometric elements in the design drawings to extract the three-dimensional spatial coordinates of each component, node, and support, constructing a complete geometric shape spatial coordinate system.

[0033] Step S223 involves fusing and comparing the structural parameters with the geometric spatial coordinates, extracting difference parameters based on the comparison results, and correcting the geometric spatial coordinates according to these difference parameters to generate three-dimensional solid spatial beam data. Specifically, the geometric parameters in the structural parameters, such as component cross-sectional dimensions and length, are fused and compared with the component length and position coordinates in the geometric spatial coordinates. Difference parameters are calculated, such as the difference between the designed length and the measured length, and the difference between the designed cross-sectional dimensions and the measured cross-sectional dimensions. The geometric spatial coordinates are corrected based on these difference parameters; for example, if the measured length is 0.05m longer than the designed length, the coordinates of the component's endpoint node are adjusted. Using the secondary development interface of the 3D modeling software, three-dimensional solid spatial beam data containing the component's cross-sectional shape, dimensions, material, and spatial position is generated based on the corrected geometric spatial coordinates and structural parameters.

[0034] Step S224: Based on the three-dimensional solid space beam and rod data, mesh generation and value assignment are performed to construct the initial structural reference model. Specifically, the three-dimensional solid space beam and rod data is meshed using tetrahedral or hexahedral elements, and the mesh size is set, such as the element length being 1 / 10 to 1 / 5 of the member length. Using the preprocessing module of the finite element analysis software, the material parameters and mechanical parameters in the structural parameters are assigned to the corresponding mesh elements and nodes. Boundary conditions of the structure are set, such as the constraint types for fixed supports and elastic supports, finally constructing an initial structural reference model that can be used for simulation calculations.

[0035] Step S300: Compare the theoretical structural response dataset with the structural response dataset, identify anomalies based on the comparison results, and locate abnormal response regions.

[0036] Specifically, the theoretical structural response dataset and the measured structural response dataset are compared from multiple dimensions such as time, space, and numerical values. The consistency between the two datasets is analyzed, and abnormal data is marked according to the differences between the two datasets. That is, data points in the two datasets whose numerical differences exceed a preset threshold, whose change trends are inconsistent, or whose mechanical laws are not followed are marked. Through spatiotemporal analysis of abnormal data, regions in the large space frame structure where the response is abnormal are identified and located. These regions are spatial regions where the structural response, such as strain and displacement, deviates significantly from the theoretical simulation results. These regions may have changes in mechanical properties due to component fatigue, node loosening, damage accumulation, etc.

[0037] In one possible implementation, based on the comparison between the theoretical structural response dataset and the structural response dataset, anomalies are identified and located according to the comparison results. Step S300 further includes step S310, aligning the theoretical structural response dataset and the structural response dataset with their timestamps to generate first aligned data. Specifically, a time synchronization algorithm is used to align the two sets of data with the same time granularity, using the timestamps of the structural response dataset and the theoretical structural response dataset as a reference. For data with inconsistent timestamps, interpolation is used to supplement missing time point data, or downsampling is used to unify the data frequency, ultimately generating first aligned data with completely consistent timestamps.

[0038] Step S320: Align the theoretical structural response dataset with the structural response dataset in three-dimensional space to generate second aligned data. Specifically, extract the spatial coordinates from the two sets of data, including the sensor installation location coordinates of the measured data and the component / node coordinates of the theoretical data. Use a spatial coordinate registration algorithm to align the spatial coordinates of the two sets of data. If there is a deviation between the measured sensor position and the node / component position of the theoretical model, map the spatial coordinates of the measured data to the coordinate system of the theoretical model through coordinate transformation, and finally generate second aligned data with completely corresponding spatial positions.

[0039] Step S330: Based on the fusion of the first aligned data and the second aligned data, a time-space analysis framework is constructed. Specifically, a multi-dimensional data index structure, such as a hash index or a tree index, is created using timestamps and three-dimensional spatial coordinates as a joint primary key to enable fast data querying and association. Based on the spatial mapping relationship of the second aligned data, the time series data of the first aligned data is filled into the multi-dimensional data index structure to form spatiotemporally related data. The spatiotemporally related data is time-sliced, and spatial distribution data is queried within each time slice. Multiple time slices are combined with the corresponding spatial distribution data in an array to construct a time-space analysis framework that can simultaneously represent the time and spatial dimensions.

[0040] Step S340: Based on the time-space analysis framework, perform point-by-point difference calculation on the theoretical structural response dataset and the structural response dataset to generate a residual data field. Specifically, based on the time-space analysis framework, perform point-by-point difference calculation on the theoretical response value and the measured response value at each spatiotemporal point, where residual = measured value - theoretical value. Arrange the residual values ​​of all spatiotemporal points according to their corresponding timestamps and three-dimensional spatial coordinates to form a residual data field containing three dimensions: time, space, and residual. The residual data field is stored using a three-dimensional array or grid data structure, which can intuitively reflect the distribution of residual values ​​in time and space.

[0041] Step S350: Perform multidimensional anomaly feature analysis based on the residual data field to determine multidimensional anomaly features, and identify multiple anomalous data points according to these features. Specifically, set multidimensional anomaly judgment thresholds such as dynamic amplitude threshold and frequency band energy threshold. The thresholds are determined through statistical methods, such as the dynamic amplitude threshold being the mean of all residual values ​​in the residual data field plus three times the standard deviation. Based on the dynamic amplitude threshold, perform time-domain amplitude analysis on the residual data field to identify time-domain amplitude anomalies where residual values ​​exceed the threshold. Based on the frequency band energy threshold, perform a fast Fourier transform on the residual time series to calculate the energy of each frequency band and identify frequency-domain energy anomalies where energy exceeds the threshold. Set a moving time window, traverse the residual data field, and calculate the statistical distribution of residual values ​​within each window, such as mean, variance, and probability density, to identify statistical distribution anomalies where the statistical distribution does not conform to normal patterns. Integrate the three types of anomalies to obtain the final multiple anomalous data points.

[0042] Step S360: Perform spatiotemporal analysis on the multiple abnormal data points, extract multiple spatiotemporally adjacent abnormal data points for spatial clustering, and delineate the response abnormal region. Specifically, perform spatiotemporal analysis on the identified multiple abnormal data points, extracting the timestamp and three-dimensional spatial coordinates of each abnormal data point. Using a spatial clustering algorithm, spatiotemporally adjacent abnormal data points with a spatial distance less than a preset clustering threshold are clustered into one class. Each cluster corresponds to a potential abnormal region. Calculate the spatial boundary of each cluster, such as the minimum bounding cube or convex hull, to delineate the range of the response abnormal region. Output the spatial coordinate range, the number of abnormal data points contained within, and the degree of abnormality for each abnormal region.

[0043] In one possible implementation, a time-space analysis framework is constructed by fusing the first alignment data and the second alignment data. Step S330 further includes step S331, creating a multidimensional data index structure using the timestamp and three-dimensional spatial coordinates as a composite primary key. Specifically, the timestamp and three-dimensional spatial coordinates are combined as a composite primary key, and a hash index algorithm is used to construct the multidimensional data index structure. The hash function uses the MD5 algorithm to convert the composite primary key into a fixed-length hash value, which serves as the index key. The index value stores pointers to the measured response data or theoretical response data at that spatiotemporal point, or directly stores the data content. An index dictionary is constructed to enable quick retrieval of corresponding data using the composite primary key.

[0044] Step S332: Define a spatial mapping relationship based on the second alignment data, and fill the first alignment data into the multidimensional data index structure according to the spatial mapping relationship. Specifically, based on the spatial mapping relationship determined in the second alignment data, such as the correspondence between sensor installation locations and theoretical model nodes, establish a spatial association table between measured data and theoretical data. Traverse each time series data in the first alignment data and find the corresponding theoretical model spatial coordinates according to the spatial association table. Fill the measured values ​​and theoretical values ​​in the time series data into the multidimensional data index structure according to the composite primary key, ensuring that the index value corresponding to each index key contains complete time series data.

[0045] Step S333: Based on the multidimensional data index structure, time slicing is performed to determine multiple time slices. Spatial queries are then performed according to these multiple time slices to determine the spatial distribution data. Specifically, the length of the time slices is set, such as 5s, 10s, and 30s, dividing the entire time series into multiple consecutive and non-overlapping time slices. For each time slice, the composite primary key corresponding to all timestamps within that time slice is extracted. The spatial coordinates and response data corresponding to these composite primary keys are queried through the multidimensional data index structure to obtain the spatial distribution data within that time slice. The spatial distribution data is organized into a spatial matrix, with rows representing spatial points and columns representing response parameters for analysis.

[0046] Step S334: Based on the multiple time slices and the spatial distribution data, an array is combined to construct the time-space analysis framework. Specifically, the spatial distribution matrices of multiple time slices are combined in time slice order to form a three-dimensional tensor, i.e., time slice dimension × spatial point dimension × response parameter dimension. The dimensions of this three-dimensional tensor are labeled, clarifying the meaning and indexing rules of each dimension, such as time slice dimension index 1 corresponding to time slice 1, spatial point dimension index 1 corresponding to spatial point 1, and response parameter dimension index 1 corresponding to strain value. A data query interface is constructed, allowing for rapid extraction of corresponding data through time slice number, spatial point number, and response parameter type, ultimately forming a time-space analysis framework capable of supporting spatiotemporal joint analysis.

[0047] In one possible implementation, the theoretical structural response dataset and the structural response dataset are traversed point-by-point according to the time-space analysis framework to calculate the difference, generating a residual data field. Step S340 further includes step S341, analyzing the theoretical structural response dataset and the structural response dataset according to a preset time period based on the time-space analysis framework, and extracting the measured response time history curve and the theoretical response time history curve. Specifically, based on the time-space analysis framework, the measured response time series and the theoretical response time series of each spatial point are extracted according to a preset time period, such as the entire analysis duration or a specific load application period. The time series data of each spatial point are arranged in timestamp order, and the measured response time history curve and the theoretical response time history curve are plotted, where the horizontal axis represents time and the vertical axis represents the response parameter values. The response time history curves are stored in the form of line graphs, and each curve contains a timestamp array and a response value array.

[0048] Step S342 involves subtracting the measured response time history curve from the theoretical response time history curve using time-by-time subtraction to generate a residual time series. Specifically, using the theoretical response time history curve as a reference, the measured response value at each time point is subtracted from the theoretical response value, i.e., residual = measured value - theoretical value. This process is repeated for all time points within the preset time period, calculating the residual value at each time point to form a residual time series, where the horizontal axis represents time and the vertical axis represents the residual value. The residual time series is also stored in the format of a timestamp array and a residual value array for easy analysis.

[0049] Step S343: Traverse the residual time series to perform spatial analysis of the measurement point source and determine the spatial coordinates of the measurement points. Specifically, traverse each residual time series and query the three-dimensional spatial coordinates of the corresponding measurement points through a multi-dimensional data index structure. Associate the three-dimensional spatial coordinates with the residual time series and store them as key-value pairs of three-dimensional spatial coordinates and residual time series.

[0050] Step S344: Traverse the residual time series to perform measurement point source time analysis and determine the measurement point time parameters. Specifically, traverse each residual time series and extract the time parameters of the series, including the start timestamp, end timestamp, time interval, and number of data points. Associate the time parameters with the residual time series and the three-dimensional spatial coordinates to form a complete data structure of three-dimensional spatial coordinates-time parameters-residual time series.

[0051] Step S345: Construct the residual data field according to the spatiotemporal dimensions based on the spatial coordinates of the measurement points and the time parameters of the measurement points. Specifically, based on the three-dimensional spatial coordinates, time parameters, and residual time series of all measurement points, a three-dimensional grid interpolation algorithm is used to interpolate the discrete residual data into a continuous three-dimensional spatial grid. The interpolated spatial grid data is combined with the time dimension according to the time parameters to form a continuous residual data field containing the three dimensions of time, space, and residuals. The residual data field is stored in a three-dimensional array or volume data format, which can support the query and visualization of residual values ​​at any spatiotemporal point.

[0052] In one possible implementation, multidimensional anomaly feature analysis is performed based on the residual data field to determine multidimensional anomaly features. Multiple anomalous data points are identified according to these multidimensional anomaly features. Step S350 further includes step S351, setting a dynamic amplitude threshold and performing temporal amplitude analysis on the residual data field to determine temporal anomaly features. Specifically, a statistical method is used to set the dynamic amplitude threshold, calculating the mean and standard deviation of all residual values ​​in the residual data field. The dynamic amplitude threshold = mean ± k × standard deviation, where k is an empirical coefficient. Each spatiotemporal point in the residual data field is traversed, and it is determined whether its residual value exceeds the dynamic amplitude threshold. The spatiotemporal points corresponding to residual values ​​exceeding the threshold are marked as temporal anomaly points. The residual values, timestamps, and spatial coordinates of these points are extracted to form a set of temporal anomaly features.

[0053] Step S352: Based on the aforementioned time-domain anomaly features, perform point comparison and determination to identify time-domain amplitude anomalies. Specifically, compare the spatiotemporal points in the time-domain anomaly feature set with the structural parameters of the theoretical model, such as component positions and node distribution, to confirm the physical component or node corresponding to each anomaly point. False anomalies caused by sensor malfunctions or data acquisition errors are eliminated. For example, if all time point data of a certain sensor exceeds the threshold and other adjacent sensors show no anomalies, it is determined to be a sensor malfunction, and the anomaly point corresponding to that sensor is eliminated. Finally, time-domain amplitude anomalies related to the actual stress on the structure are retained.

[0054] Step S353: Set a frequency band energy threshold and perform frequency domain energy analysis by traversing the residual data field to determine frequency band energy anomaly characteristics. Specifically, to set a frequency band energy threshold, first, perform a Fast Fourier Transform on all residual time series in the residual data field to convert the time domain signal into a frequency domain signal. Calculate the energy proportion of each frequency band, i.e., energy of a certain frequency band / total energy × 100%. Based on the inherent frequency characteristics of the structure, set an energy threshold for key frequency bands, such as the frequency band where the structure's inherent frequency is located. Traverse all residual time series and determine whether the energy proportion of its key frequency band is lower than the threshold. Mark the spatiotemporal points corresponding to residual time series that are lower than the threshold as frequency domain energy anomaly points. Extract the frequency domain features of these points, such as the energy proportion of each frequency band and spectral peak values, to form a set of frequency band energy anomaly features.

[0055] Step S354: Based on the frequency band energy anomaly characteristics, perform point comparison and determination to identify frequency domain energy anomaly points. Specifically, compare the spatiotemporal points in the frequency band energy anomaly feature set with the structural parameters of the theoretical model to confirm the physical component or node corresponding to each anomaly point. Combined with the structure's dynamic characteristics, such as natural frequencies and mode shapes, analyze the causes of frequency domain energy anomalies; for example, structural damage may cause changes in natural frequencies, thus affecting the frequency domain energy distribution. False anomalies caused by external interference such as environmental noise and equipment vibration are eliminated, ultimately retaining the frequency domain energy anomalies related to structural damage.

[0056] Step S355: Perform statistical analysis on the residual data field by traversing the moving time window to determine the probability distribution characteristics. Specifically, set a moving time window, such as a window size of 20-50 time points, traverse each residual time series in the residual data field, perform statistical analysis on the residual values ​​within each moving window, calculate statistics such as mean, variance, standard deviation, skewness, and kurtosis, and construct probability distribution characteristics. Based on the normal structure of the residual probability distribution, set statistical thresholds, determine whether the statistics within each moving window exceed the thresholds, and mark the spatiotemporal points corresponding to moving windows that exceed the thresholds as statistical distribution outliers, forming a set of probability distribution characteristics.

[0057] Step S356: Based on the probability distribution characteristics, perform point comparison and judgment to identify statistical distribution anomalies. Specifically, compare the spatiotemporal points in the probability distribution feature set with the structural parameters of the theoretical model to confirm the physical component or node corresponding to each anomaly. Analyze the continuity of statistical distribution anomalies, eliminate isolated false anomalies, such as when the statistic of a single spatiotemporal point exceeds a threshold while surrounding points are not abnormal, and finally retain statistical distribution anomalies with continuity.

[0058] Step S357: Integrate the time-domain amplitude anomalies, frequency-domain energy anomalies, and statistical distribution anomalies to obtain the multiple anomalous data points. Specifically, set operations are used to integrate the time-domain amplitude anomalies, frequency-domain energy anomalies, and statistical distribution anomalies. If a union is used, all three types of anomalies are retained, and duplicates are removed; if an intersection is used, only points belonging to all three types of anomalies are retained. Finally, the integrated multiple anomalous data points are obtained, and the timestamp, spatial coordinates, and anomaly type of each anomaly point are output.

[0059] Step S400: Based on the real-time load data, perform evolution analysis to construct time-series change characteristics. Based on the response anomaly region and the time-series change characteristics, perform fatigue damage evolution on the large space frame structure and generate a fatigue damage detection report.

[0060] Specifically, time series analysis is performed on real-time load data to extract the temporal variation characteristics of the load, i.e., the variation law of the load in the time dimension, such as the number of load cycles, load amplitude level, cycle frequency, and duration. Combined with the identified abnormal response areas, the fatigue damage development process in these areas is analyzed, i.e., the process by which damage such as crack initiation and propagation gradually accumulates over time in structural components under alternating loads, such as the damage accumulation rate and damage degree. Finally, a fatigue damage detection report containing information such as the location of the abnormal area, fatigue damage level, and remaining life is generated.

[0061] In one possible implementation, evolution analysis is performed based on the real-time load data to construct time-series change characteristics. Step S400 further includes step S410, performing polymorphic separation and identification based on the real-time load data to obtain static load components and dynamic load fluctuation components. Specifically, a signal decomposition algorithm, such as wavelet decomposition, is used to decompose the real-time load data and separate the static load components and dynamic load fluctuation components. The static load components are low-frequency components in the load data, such as structural self-weight and fixed equipment weight; the dynamic load fluctuation components are high-frequency components in the load data, such as wind load and vibration load. The static load components are extracted using a low-pass filter, and the dynamic load fluctuation components are extracted using a high-pass filter.

[0062] Step S420: Based on the dynamic load fluctuation components, perform cyclic counting according to the load cycle to generate the load cycle count and load cycle occurrence frequency. Specifically, the rainflow counting method is used to perform cyclic counting on the dynamic load fluctuation components. The time series of the dynamic load fluctuation components is converted into a peak-valley sequence. According to the rainflow counting rules, i.e., starting from each peak or valley, counting downwards or upwards until a larger peak or smaller valley is encountered, the load cycle count is counted. At the same time, the occurrence frequency of each load cycle is calculated, i.e., cycle count / time length, and finally, statistical data on the load cycle count and load cycle occurrence frequency are generated.

[0063] Step S430: Amplitude analysis is performed on the dynamic load fluctuation components to divide the load into multiple amplitude levels. The load cycle count and cycle frequency are then analyzed according to these amplitude levels to construct a load amplitude-frequency distribution spectrum. Specifically, amplitude levels are set using an equal-interval division method based on the amplitude range of the dynamic load fluctuation components. For each amplitude level, the corresponding load cycle count and cycle frequency are statistically analyzed. A load amplitude-frequency distribution spectrum is constructed with the amplitude level as the horizontal axis and the cycle count and cycle frequency as the vertical axis. The distribution spectrum is stored in a two-dimensional table or line graph format to display the cycle frequency distribution pattern of loads with different amplitudes.

[0064] Step S440: Based on the load amplitude-frequency distribution spectrum, perform statistical analysis to extract key load characteristic indicators. Specifically, perform statistical analysis on the load amplitude-frequency distribution spectrum to extract key characteristic indicators that characterize the load cyclic properties, including maximum load amplitude, cumulative number of cycles, equivalent root mean square amplitude, and load amplitude distribution entropy. The maximum load amplitude is the upper limit of the maximum amplitude range in the distribution spectrum; the cumulative number of cycles is the sum of the number of cycles for all amplitude levels; the equivalent root mean square amplitude is obtained through the formula... The load amplitude distribution entropy is calculated using the information entropy formula, characterizing the uniformity of the load amplitude distribution. This ultimately forms a set of key load characteristic indicators that include the aforementioned metrics.

[0065] Step S450: Track the changes of the key load characteristic indicators in chronological order to construct the temporal change characteristics. Specifically, with time as the horizontal axis, the key load characteristic indicators are segmented and statistically analyzed according to the monitoring time sequence. The maximum load amplitude, cumulative cycle count, and equivalent root mean square amplitude, etc., are calculated for each time period to form a temporal characteristic sequence. Trend fitting algorithms, such as linear fitting and polynomial fitting, are used to analyze the temporal change trends of each indicator and identify abrupt changes in load characteristics. The temporal characteristic sequence and trend fitting results are integrated to construct a temporal change characteristic that reflects the evolution of the load over time.

[0066] In one possible implementation, fatigue damage evolution of a large space frame structure is performed based on the abnormal response region and the temporal variation characteristics to generate a fatigue damage detection report. Step S400 further includes step S460, which involves performing long-term field measurements based on the abnormal response region to extract an abnormal response dataset. Specifically, high-density long-term monitoring sensors are deployed for the defined abnormal response region. A data acquisition module is used to collect real-time response data such as strain and displacement of components within the abnormal region. Data filtering rules are set to remove outliers caused by sensor drift. The collected response data is structured and stored according to timestamps and spatial locations. Response data fragments related to the temporal variation characteristics of the load are extracted, such as response data corresponding to periods of sudden load changes, to form an abnormal response dataset.

[0067] Step S470: The abnormal response dataset is correlated with the time-series variation characteristics to generate correlation factors. Specifically, a grey relational analysis algorithm is used, with the abnormal response dataset as a reference sequence and the load time-series variation characteristics as a comparison sequence. The correlation coefficient between the reference sequence and the comparison sequence is calculated. A weighted average of the correlation coefficients is then performed to obtain the correlation degree between different load characteristic indicators and the abnormal response. The correlation degree value corresponding to the indicator with a correlation degree greater than a preset correlation degree threshold is defined as the correlation factor, characterizing the degree of correlation between load change and abnormal response.

[0068] Step S480: Perform local response analysis on the abnormal response region according to the correlation factors to generate local response patterns. Based on these patterns, perform fatigue damage deduction and construct a local stress spectrum. Specifically, based on the correlation factors, select the load characteristic indicators that have the greatest impact on the abnormal response, such as the equivalent root mean square amplitude. Use the stress-strain conversion formula to convert the strain values ​​in the abnormal response dataset into stress values. Combined with the load time-series variation characteristics, use the rainflow counting method to perform cyclic counting of the local stress data, counting the number of cycles for different stress amplitudes. Construct a local stress spectrum reflecting the stress cycle characteristics of the abnormal region, with stress amplitude on the horizontal axis and the number of cycles on the vertical axis.

[0069] Step S490: Based on the local stress spectrum, fatigue accumulation analysis is performed to calculate the cumulative fatigue degree and damage evolution, generating remaining fatigue life data. Specifically, Miner's linear cumulative damage theory is used to calculate the cumulative fatigue degree of the component in the abnormal region, using the following formula: Where, n i N represents the number of cycles for the i-th stress amplitude. i The fatigue life corresponding to the i-th stress amplitude can be obtained from the material's SN curve. When the cumulative fatigue degree D reaches 1, the component is considered to have failed due to fatigue. Based on the current cumulative fatigue degree and the trend fitting results of the load time sequence variation characteristics, the remaining time for the component to reach fatigue failure, i.e., the remaining fatigue life, is predicted.

[0070] Step S4100: The remaining fatigue life data is spatially distributed and mapped according to the abnormal response regions to construct the fatigue damage detection report. Specifically, based on the remaining fatigue life data, combined with information such as the spatial coordinates of the abnormal response regions, load temporal variation characteristics, and local stress spectra, a fatigue damage detection report is constructed according to a standardized report format. The report content includes basic structural information, monitoring plan, abnormal region identification results, load characteristic analysis, fatigue damage assessment, and maintenance recommendations. The remaining fatigue life data is spatially distributed and mapped according to the abnormal response regions, and the remaining life distribution of different regions is displayed in the form of a heat map, intuitively presenting the fatigue damage state of the structure.

[0071] This application employs various techniques, including deploying multiple types of sensors on a large space frame structure to synchronously collect response data to obtain a structural response dataset, simultaneously acquiring real-time load data and constructing a digital twin model based on structural parameters, obtaining a theoretical structural response dataset through simulation calculations, comparing the two datasets and identifying anomalies, locating abnormal response areas, performing evolution analysis on real-time load data to construct temporal change characteristics, combining the abnormal response areas to perform fatigue damage evolution analysis, and generating a fatigue damage detection report. These techniques solve the technical problem of insufficient accuracy in damage location and assessment in existing fatigue damage detection methods for large space frame structures, achieving the technical effect of improving the accuracy of damage location and assessment.

[0072] In the above text, refer to Figure 1 A fatigue damage detection method for large space frame structures according to embodiments of the present invention is described in detail. Next, reference will be made to... Figure 2 A fatigue damage detection system for large space frame structures according to an embodiment of the present invention is described.

[0073] The fatigue damage detection system for large space frame structures according to embodiments of the present invention addresses the technical problem of insufficient accuracy in damage localization and assessment in existing fatigue damage detection methods for large space frame structures, thereby improving the accuracy of damage localization and assessment. The fatigue damage detection system for large space frame structures includes: a structural response data acquisition module 10, a space frame structure digital twin model construction module 20, an anomaly comparison module 30, and a fatigue damage evolution module 40.

[0074] The structural response data acquisition module 10 is used to synchronously acquire responses from multiple types of sensors deployed on a large space frame structure to obtain a structural response dataset. The space frame structure digital twin model construction module 20 is used to synchronously acquire real-time load data of the large space frame structure and combine it with the structural parameters of the large space frame structure to construct a digital twin model of the space frame structure. Simulation calculations are performed using the space frame structure digital twin model to obtain a theoretical structural response dataset. The anomaly comparison module 30 is used to compare the theoretical structural response dataset with the structural response dataset, identify anomalies based on the comparison results, and locate abnormal response areas. The fatigue damage evolution module 40 is used to perform evolution analysis based on the real-time load data, construct time-series change characteristics, and perform fatigue damage evolution on the large space frame structure based on the abnormal response areas and the time-series change characteristics to generate a fatigue damage detection report.

[0075] The detailed description of the specific configuration of the digital twin model construction module 20 for the space frame structure is explained as follows: As mentioned above, the real-time load data of the large space frame structure is synchronously acquired and combined with the structural parameters of the large space frame structure to construct a digital twin model of the space frame structure. Simulation calculations are performed using the digital twin model of the space frame structure to obtain a theoretical structural response dataset. The digital twin model construction module 20 for the space frame structure may further include: a real-time monitoring data stream access unit for synchronously accessing the real-time monitoring data stream of the large space frame structure, using the real-time monitoring data stream as a dynamic external load input; and a structure-load interaction analysis unit for mapping the dynamic external load input to an initial structural reference model. Based on the initial structural baseline model, a structure-load interaction analysis is performed to generate dynamic structure-load interaction state parameters. A digital twin model construction unit for the space frame structure is used to construct a digital twin model of the space frame structure based on the dynamic structure-load interaction state parameters and the structural parameters. A real-time simulation loop unit is used to drive the digital twin model of the space frame structure to perform real-time simulation loops at multiple time steps to obtain a mechanical equilibrium dataset. A multi-dimensional feature analysis unit is used to perform multi-dimensional feature analysis based on the mechanical equilibrium dataset to determine multi-dimensional simulation feature values. A data alignment unit is used to align the multi-dimensional simulation feature values ​​according to the structural parameters to construct a theoretical structural response dataset.

[0076] The initial structural baseline model construction process, specifically the structure-load interaction analysis unit, may further include: a geometric topology relationship construction subunit for retrieving engineering archives of large-scale space frame structures, traversing the archives for geometric analysis, and constructing geometric topology relationships; a three-dimensional space scanning subunit for performing three-dimensional space scanning based on the geometric topology relationships, and constructing geometric shape space coordinates; a correction subunit for fusing and comparing the structural parameters with the geometric shape space coordinates, extracting difference parameters based on the comparison results, and correcting the geometric shape space coordinates according to the difference parameters to generate three-dimensional solid space beam and rod data; and a mesh generation and assignment subunit for performing mesh generation and assignment based on the three-dimensional solid space beam and rod data to construct the initial structural baseline model.

[0077] The specific configuration of the anomaly comparison module 30 is described in detail below: As mentioned above, based on the comparison between the theoretical structural response dataset and the structural response dataset, anomaly identification is performed according to the comparison results to identify and locate response anomaly regions. The anomaly comparison module 30 may further include: a timestamp alignment unit for aligning the theoretical structural response dataset with the structural response dataset using timestamps to generate first aligned data; a three-dimensional spatial alignment unit for aligning the theoretical structural response dataset with the structural response dataset in three-dimensional space to generate second aligned data; and a data fusion unit for aligning the theoretical structural response dataset with the structural response dataset based on the first aligned data. The second aligned data is fused to construct a time-space analysis framework; the point-by-point traversal difference calculation unit is used to perform point-by-point traversal difference calculation on the theoretical structural response dataset and the structural response dataset according to the time-space analysis framework to generate a residual data field; the multidimensional anomaly feature analysis unit is used to perform multidimensional anomaly feature analysis based on the residual data field, determine multidimensional anomaly features, and identify multiple anomaly data points according to the multidimensional anomaly features; the response anomaly region delineation unit is used to traverse the multiple anomaly data points for spatiotemporal analysis, extract multiple spatiotemporally adjacent anomaly data points for spatial clustering, and delineate the response anomaly region.

[0078] The time-space analysis framework is constructed by fusing the first alignment data and the second alignment data. The data fusion unit may further include: a multidimensional data index structure creation subunit for creating a multidimensional data index structure using timestamps and three-dimensional spatial coordinates as a joint primary key; a spatial mapping relationship definition subunit for defining a spatial mapping relationship based on the second alignment data and filling the first alignment data into the multidimensional data index structure according to the spatial mapping relationship; a time slicing subunit for performing time slicing based on the multidimensional data index structure to determine multiple time slices, performing spatial queries according to the multiple time slices, and determining spatial distribution data; and an array combination subunit for combining the multiple time slices and the spatial distribution data into an array to construct the time-space analysis framework.

[0079] The process involves calculating the difference between the theoretical structural response dataset and the structural response dataset using a point-by-point traversal based on the time-space analysis framework to generate a residual data field. The point-by-point difference calculation unit may further include: a response time history curve extraction subunit, used to analyze the theoretical structural response dataset and the structural response dataset according to a preset time period based on the time-space analysis framework, extracting the measured response time history curve and the theoretical response time history curve; a time-point subtraction operation subunit, used to perform time-point subtraction on the measured response time history curve using the theoretical response time history curve as a benchmark, generating a residual time series sequence; a measurement point source space analysis subunit, used to traverse the residual time series sequence to perform measurement point source space analysis and determine the measurement point spatial coordinates; a measurement point source time analysis subunit, used to traverse the residual time series sequence to perform measurement point source time analysis and determine the measurement point time parameters; and a residual data field construction subunit, used to construct the residual data field based on the measurement point spatial coordinates and the measurement point time parameters according to the spatiotemporal dimension.

[0080] The multidimensional anomaly feature analysis unit, based on the residual data field, performs multidimensional anomaly feature analysis to determine multidimensional anomaly features and identifies multiple anomalous data points according to these features. The multidimensional anomaly feature analysis unit may further include: a time-domain amplitude analysis subunit for setting a dynamic amplitude threshold and traversing the residual data field to perform time-domain amplitude analysis to determine time-domain anomaly features; a first point comparison and determination subunit for performing point comparison and determination based on the time-domain anomaly features to identify time-domain amplitude anomaly points; and a frequency-domain energy analysis subunit for setting a frequency band energy threshold and traversing the residual data field to perform frequency-domain energy analysis. The system employs a three-tiered approach: a first tier determines frequency band energy anomaly characteristics; a second tier performs point comparison and determination based on these characteristics to identify frequency domain energy anomalies; a third tier performs statistical analysis by traversing the residual data field according to a moving time window to determine probability distribution characteristics; and an anomaly integration tier integrates the time domain amplitude anomalies, frequency domain energy anomalies, and statistical distribution anomalies to obtain the multiple anomaly data points.

[0081] The fatigue damage evolution module 40 is described in detail below: As mentioned above, based on the real-time load data, evolution analysis is performed to construct time-series change characteristics. The fatigue damage evolution module 40 may further include: a multi-state separation and identification unit for performing multi-state separation and identification based on the real-time load data to obtain static load components and dynamic load fluctuation components; a cycle counting unit for performing cycle counting based on the dynamic load fluctuation components according to the load cycle to generate the number of load cycles and the frequency of load cycles; an amplitude analysis unit for performing amplitude analysis based on the dynamic load fluctuation components, dividing multiple amplitude levels, and analyzing the number of load cycles and the frequency of load cycles according to the multiple amplitude levels to construct a load amplitude-frequency distribution spectrum; a load key feature index extraction unit for performing statistics based on the load amplitude-frequency distribution spectrum to extract load key feature indicators; and a change tracking unit for performing change tracking according to the load key feature indicators in chronological order to construct the time-series change characteristics.

[0082] The fatigue damage evolution module 40, which performs fatigue damage evolution analysis on the large-scale space frame structure based on the abnormal response region and the temporal variation characteristics, and generates a fatigue damage detection report, may further include: a long-cycle measurement unit for performing long-cycle measurements based on the abnormal response region to extract abnormal response datasets; a correlation analysis unit for performing correlation analysis between the abnormal response datasets and the temporal variation characteristics to generate correlation factors; a local response analysis unit for performing local response analysis on the abnormal response region according to the correlation factors to generate local response patterns, and performing fatigue damage deduction based on the local response patterns to construct a local stress spectrum; a fatigue accumulation analysis unit for performing fatigue accumulation analysis based on the local stress spectrum, calculating cumulative fatigue degree to perform damage evolution, and generating remaining fatigue life data; and a spatial distribution mapping unit for performing spatial distribution mapping on the remaining fatigue life data according to the abnormal response region to construct the fatigue damage detection report.

[0083] The fatigue damage detection system for large space frame structures provided in this embodiment of the invention can execute the fatigue damage detection method for large space frame structures provided in any embodiment of the invention, and has the corresponding functional modules and beneficial effects of the method.

[0084] Although this application makes various references to certain modules in the system according to the embodiments of this application, any number of different modules can be used and run on user terminals and / or servers. The various units and modules included are only divided according to functional logic, but are not limited to the above division, as long as the corresponding functions can be achieved; in addition, the specific names of each functional unit are only for easy distinction between each other and are not used to limit the scope of protection of this invention.

[0085] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any modifications, equivalent changes, and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A method for detecting fatigue damage of a large-scale space truss structure, characterized by, The method comprises: Based on the large grid structure layout of multiple types of sensors for synchronous response acquisition, obtain the structure response data set; Synchronously acquire the real-time load data of the large grid structure, combine the structure parameters of the large grid structure to construct the grid structure digital twin model, and perform simulation calculation through the grid structure digital twin model to obtain the theoretical structure response data set; Based on the comparison between the theoretical structure response data set and the structure response data set, abnormal identification is performed according to the comparison result, and the response abnormal area is identified and located; Based on the real-time load data, the time sequence change characteristics are constructed, the fatigue damage evolution of the large grid structure is performed based on the response abnormal area and the time sequence change characteristics, and a fatigue damage detection report is generated.

2. The method for detecting fatigue damage of a large space truss structure according to claim 1, wherein Synchronously acquire the real-time load data of the large grid structure, combine the structure parameters of the large grid structure to construct the grid structure digital twin model, and perform simulation calculation through the grid structure digital twin model to obtain the theoretical structure response data set, the method comprising: Synchronously access the real-time monitoring data stream of the large grid structure, and input the real-time monitoring data stream as a dynamic external load; Map the dynamic external load input to an initial structure reference model, perform structure-load interaction analysis based on the initial structure reference model, and generate dynamic structure-load interaction state parameters; Based on the dynamic structure-load interaction state parameters and the structure parameters, a grid structure digital twin model is constructed; Drive the grid structure digital twin model to perform real-time simulation circulation according to multiple time steps to obtain a mechanical equilibrium data set; Based on the mechanical equilibrium data set, multi-dimensional feature analysis is performed to determine multi-dimensional simulation characteristic values; Align the multi-dimensional simulation characteristic values according to the structure parameters to construct a theoretical structure response data set.

3. The method for detecting fatigue damage of a large space truss structure according to claim 2, wherein The construction process of the initial structure reference model comprises: Access the engineering archives of the large grid structure, traverse the engineering archives for geometric analysis, and construct geometric topological relationship; Based on the geometric topological relationship, perform three-dimensional space scanning to construct geometric morphological space coordinates; Fuse and compare the structure parameters and the geometric morphological space coordinates, extract difference parameters according to the comparison result, modify the geometric morphological space coordinates according to the difference parameters, and generate three-dimensional entity space beam data; Based on the three-dimensional entity space beam data, perform grid division and assignment to construct the initial structure reference model.

4. The method for detecting fatigue damage of a large space truss structure according to claim 1, wherein Based on the comparison between the theoretical structure response data set and the structure response data set, abnormal identification is performed according to the comparison result, and the response abnormal area is identified and located, the method comprising: Align the theoretical structure response data set and the structure response data set by time stamp to generate first alignment data; Align the theoretical structure response data set and the structure response data set in three-dimensional space to generate second alignment data; Fuse the first alignment data and the second alignment data to construct a time-space analysis framework; According to the time-space analysis framework, point-by-point traversal difference calculation is performed on the theoretical structure response data set and the structure response data set to generate residual data field; Performing multi-dimensional abnormal feature analysis based on the residual data field, determining multi-dimensional abnormal features, and judging according to the multi-dimensional abnormal features to identify multiple abnormal data points; Performing time-space analysis by traversing the multiple abnormal data points, extracting multiple time-space adjacent abnormal data points for spatial clustering, and demarcating a response abnormal area.

5. The method for detecting fatigue damage of a large space truss structure according to claim 4, wherein Based on the first alignment data and the second alignment data, a time-space analysis framework is constructed, and the method comprises: Creating a multi-dimensional data index structure with time stamp and three-dimensional spatial coordinates as joint primary keys; Defining a spatial mapping relationship based on the second alignment data, and filling the first alignment data into the multi-dimensional data index structure according to the spatial mapping relationship; Based on the multi-dimensional data index structure, time slicing is performed to determine multiple time slices, and spatial query is performed according to the multiple time slices to determine spatial distribution data; Based on the multiple time slices and the spatial distribution data, an array combination is performed to construct the time-space analysis framework.

6. The method for detecting fatigue damage of a large space truss structure according to claim 4, wherein According to the time-space analysis framework, point-by-point traversal difference calculation is performed on the theoretical structure response data set and the structure response data set to generate a residual data field, and the method comprises: Based on the time-space analysis framework, the theoretical structure response data set and the structure response data set are analyzed according to a preset time period, and the measured response time history curve and the theoretical response time history curve are extracted; The theoretical response time history curve is used as a reference to perform time point subtraction operation on the measured response time history curve to generate a residual time sequence; Traversing the residual time sequence to perform source space analysis of the measuring points to determine the spatial coordinates of the measuring points; Traversing the residual time sequence to perform source time analysis of the measuring points to determine the time parameters of the measuring points; Based on the spatial coordinates of the measuring points and the time parameters of the measuring points, the residual data field is constructed according to the time-space dimensions.

7. The method for detecting fatigue damage of a large space truss structure according to claim 4, wherein Based on the residual data field, multi-dimensional abnormal feature analysis is performed to determine multi-dimensional abnormal features, and judgment is made according to the multi-dimensional abnormal features to identify multiple abnormal data points, and the method comprises: Setting a dynamic amplitude threshold to traverse the residual data field to perform time domain amplitude analysis and determine time domain abnormal features; Based on the time domain abnormal features, point position comparison and judgment are performed to determine time domain amplitude abnormal points; Setting a frequency band energy threshold to traverse the residual data field to perform frequency band energy analysis and determine frequency band energy abnormal features; Based on the frequency band energy abnormal features, point position comparison and judgment are performed to determine frequency band energy abnormal points; According to a moving time window, the residual data field is traversed for statistical analysis to determine probability distribution features; Based on the probability distribution features, point position comparison and judgment are performed to determine statistical distribution abnormal points; The time domain amplitude abnormal points, the frequency band energy abnormal points, and the statistical distribution abnormal points are integrated to obtain the multiple abnormal data points.

8. The method for detecting fatigue damage of a large space truss structure according to claim 1, wherein Based on the real-time load data, evolution analysis is performed to construct time sequence change features, and the method comprises: Based on the real-time load data, multi-state separation identification is performed to obtain static load components and dynamic load fluctuation components; Based on the dynamic load fluctuation components, cycle counting is performed according to the load period to generate load cycle number and load cycle occurrence frequency; According to the dynamic load fluctuation component, amplitude analysis is performed, a plurality of amplitude levels are divided, the load cycle number and the load cycle frequency are analyzed according to the plurality of amplitude levels, and a load amplitude-frequency distribution spectrum is constructed; Based on the load amplitude-frequency distribution spectrum, load key characteristic indexes are extracted; According to the load key characteristic indexes, time sequence change characteristics are constructed.

9. The method for detecting fatigue damage of a large space truss structure according to claim 1, wherein Based on the response abnormal area and the time sequence change characteristics, fatigue damage evolution of the large grid structure is performed, and a fatigue damage detection report is generated, the method comprising: Based on the response abnormal area, long-period measurement is performed, and an abnormal response data set is extracted; The abnormal response data set and the time sequence change characteristics are associated and analyzed, and an association factor is generated; According to the association factor, local response analysis is performed on the response abnormal area, a local response law is generated, fatigue damage is deduced according to the local response law, and a local stress spectrum is constructed; Based on the local stress spectrum, fatigue accumulation analysis is performed, cumulative fatigue degree is calculated, damage evolution is performed, and residual fatigue life data is generated; The residual fatigue life data is spatially distributed and mapped according to the response abnormal area, and the fatigue damage detection report is constructed.

10. A fatigue damage detection system for large grid structures, characterized by, The system is used to implement the fatigue damage detection method for the large grid structure according to any one of claims 1-9, and the system comprises: A structure response data acquisition module is used to acquire structure response data sets by synchronously collecting responses based on a plurality of sensors arranged on a large grid structure; A grid structure digital twin model construction module is used to construct a grid structure digital twin model by synchronously acquiring real-time load data of a large grid structure and combining structure parameters of the large grid structure, and to obtain theoretical structure response data sets by simulation calculation of the grid structure digital twin model; An abnormal comparison module is used to compare the theoretical structure response data sets with the structure response data sets, to identify abnormal areas according to comparison results, and to identify and locate the response abnormal areas; A fatigue damage evolution module is used to perform evolution analysis based on the real-time load data, to construct time sequence change characteristics, to perform fatigue damage evolution of the large grid structure based on the response abnormal area and the time sequence change characteristics, and to generate a fatigue damage detection report.

Images