Steel structure net rack intelligent detection system and method based on digital twinning

Abstract

This invention relates to the field of space frame inspection technology and discloses an intelligent inspection system and method for steel structure space frames based on digital twins. This system addresses the problem of low reliability of virtual structural models during space frame inspection. The method includes: acquiring model consistency assessment data of an initial virtual space frame model; evaluating the model effectiveness index of the virtual space frame model based on the model consistency assessment data; determining whether the initial virtual space frame model meets simulation conditions based on the model effectiveness index; if the initial virtual space frame model does not meet the simulation conditions, initiating data assimilation-based model self-correction to obtain a corrected virtual space frame model; acquiring the model effectiveness index again; terminating the data assimilation-based model self-correction when the model effectiveness index meets the simulation conditions; obtaining the final virtual space frame model; and transmitting the final virtual space frame model to the evaluation and decision-making module. This effectively reduces the uncertainty in the structural safety management and maintenance decision-making process.

Description

Technical Field

[0001] This invention relates to the field of space frame inspection technology, and more specifically to an intelligent inspection system and method for steel structure space frames based on digital twins. Background Technology

[0002] With the continuous development of building structural health monitoring and information modeling technologies, steel space frames, as a structural system with excellent spatial load-bearing performance, large span, and flexible shape, are widely used in public buildings such as stadiums, airport terminals, high-speed rail stations, and large industrial plants. During long-term service, steel space frames are continuously subjected to various external forces such as wind loads, temperature stresses, fatigue loads, and environmental corrosion. Their structural components may gradually develop various structural anomalies, including member bending, weld cracking, loose connections, coating peeling, and corrosion. If these anomalies are not detected and addressed in a timely manner, they may affect the overall load-bearing performance and service safety of the structure, and in severe cases, may even lead to structural safety risks. Therefore, conducting efficient and high-precision condition monitoring and safety assessment of steel space frames has become an important research direction in the field of structural health management.

[0003] In existing technologies, the condition monitoring of steel space frames typically relies on manual inspections or the acquisition of surface images, thermal images, or point cloud data from drones equipped with inspection equipment. Data processing and defect identification algorithms are then used to identify visible defects such as cracks, corrosion, or significant deformation. After obtaining the inspection results, the identified defect information is usually mapped onto a pre-established three-dimensional structural model or digital model. Mechanical analysis or structural simulation methods are then used to assess the impact of these defects on structural safety. This type of method has been applied in engineering practice; its basic idea is to detect structural anomalies through inspection and then conduct subsequent safety assessments and maintenance decisions based on this anomaly information.

[0004] In existing intelligent inspection processes for steel space frames, it is typically assumed that the structure is in normal operating condition when no obvious defects are identified in the inspection data. The corresponding virtual structural model, whether established or updated, accurately reflects the actual state of the physical structure. Therefore, after defect identification, the system often directly performs subsequent mechanical analysis and safety assessment based on the current virtual model. This approach has advantages in terms of system implementation complexity and data processing efficiency, and it also facilitates unified structural safety analysis, thus it is widely adopted in engineering practice.

[0005] However, the above-mentioned technologies have at least the following technical problems: During the actual service of steel space frame structures, even if no obvious defects such as cracks, corrosion, or geometric deformation are detected, the structure may still deviate from its initial design state due to factors such as long-term load effects, temperature stress changes, changes in connection node stiffness, or differences in historical operating conditions. Such changes in state may not necessarily manifest as directly identifiable defects in the inspection data, but they may have a potential impact on the overall structural performance and stability. However, existing inspection and evaluation processes typically do not differentiate the reliability of the current state of the virtual structural model, but rather assume that the absence of detected defects indicates a reliable model state. In this case, directly conducting structural analysis and safety assessments based on this virtual model may lead to deviations in the assessment results from the actual structural state, thereby increasing the uncertainty in structural safety management and maintenance decisions. Summary of the Invention

[0006] In order to overcome the above-mentioned defects of the prior art, the present invention provides a smart detection system and method for steel structure space frame based on digital twin, so as to solve the problems existing in the background art.

[0007] To achieve the above objectives, the present invention provides the following technical solution: A digital twin-based intelligent inspection system for steel structure space frames includes: a defect identification module for automated inspection of the steel structure space frame using sensors to collect raw space frame inspection data; a virtual mapping module for receiving the raw space frame inspection data and constructing an initial virtual space frame model corresponding one-to-one with the physical entity of the steel structure space frame, the virtual space frame model representing the geometric state, spatial topology, and structural operation parameters of the steel structure space frame; a reliability assessment module for acquiring model consistency assessment data of the initial virtual space frame model, including the set of all nodes in the model, the timestamp sequences of each sensor, the set of operation and maintenance event records, and the set of model update timestamps, evaluating the model effectiveness index of the virtual space frame model based on the model consistency assessment data, and determining whether the initial virtual space frame model meets the simulation conditions based on the model effectiveness index; and a model self-calibration module. The positive module, if it determines that the initial virtual space frame model does not meet the simulation conditions, initiates data assimilation-based model self-calibration to obtain a calibrated virtual space frame model. It then obtains the model effectiveness index again, and terminates the data assimilation-based model self-calibration when the model effectiveness index meets the simulation conditions, resulting in the final virtual space frame model, which is then transmitted to the evaluation and decision module. If it determines that the initial virtual space frame model meets the simulation conditions, it directly uses the initial virtual space frame model as the final virtual space frame model. The data processing module preprocesses, enhances, and extracts features from the original space frame inspection data to obtain processed space frame data. The machine learning recognition module analyzes the processed space frame data using a defect recognition algorithm model to identify defect information in the steel structure space frame. The evaluation and decision module maps the identified defect information to the final virtual space frame model, performs mechanical simulation and safety assessment, and generates maintenance decision recommendations.

[0008] Preferably, the steps for obtaining the initial virtual space frame model are as follows: receiving the original space frame inspection data collected by the defect identification module, and organizing the original space frame inspection data according to the collection time, spatial location, and data type to form an original inspection data set; based on the original inspection data set, extracting structural feature information such as the outline of the members, the position of the nodes, and the boundary of the components in the steel structure space frame to obtain structural feature data for describing the geometric shape and connection relationship of the space frame; determining the connection relationship between each member and node in the steel structure space frame according to the structural feature data, and constructing the spatial topology of the space frame; determining the geometric structural state parameters of each member in the steel structure space frame based on the structural feature data and the spatial topology, including spatial position, length direction, and node coordinates; based on the construction of the geometric structural state parameters and the spatial topology relationship, associating and assigning values ​​to the state parameters related to the operating state of the steel structure space frame with the corresponding members or nodes to form an initial virtual space frame model containing geometric structural state, spatial topology relationship, and structural operating state parameters.

[0009] Preferably, the steps for obtaining the model effectiveness index are as follows: based on the spatial topology of the initial virtual network model, obtain the set of all nodes in the virtual network model, and evaluate the structural connection verifiability coefficient based on the set of all nodes; obtain the timestamp sequence in the original network detection data corresponding to each detection sensor in the current inspection cycle, and evaluate the multi-source observation time series consistency coefficient based on each timestamp sequence; obtain the set of operation and maintenance event records in the current inspection cycle and the model update timestamp set of the initial virtual network model, and evaluate the model operation and maintenance consistency coefficient based on the set of operation and maintenance event records and the model update timestamp set; normalize the structural connection verifiability coefficient, the multi-source observation time series consistency coefficient, and the model operation and maintenance consistency coefficient, and calculate the model effectiveness index based on the normalized structural connection verifiability coefficient, the multi-source observation time series consistency coefficient, and the model operation and maintenance consistency coefficient.

[0010] Preferably, the steps for obtaining the structural connection verifiability coefficient are as follows: based on the spatial topology of the initial virtual space frame model, obtain the set of all nodes in the virtual space frame model, and select the set of key nodes from the set of all nodes; for each key node in the set of key nodes, obtain the set of connecting members corresponding to the key node based on the initial virtual space frame model, and determine the node connection constraint of the key node accordingly; extract the observation-side node connection feature data that corresponds one-to-one with the set of key nodes based on the original space frame detection data; for each key node, compare the node connection constraint of the key node with its corresponding observation-side node connection feature data one-to-one, and calculate the node connection constraint residual vector; calculate the norm value of the node connection constraint residual vector of each key node based on the node connection constraint residual vector, and calculate the node connection verifiability of each key node based on the norm value of the node connection constraint residual vector of each key node; aggregate the node connection verifiability of all key nodes in the set of key nodes to obtain the structural connection verifiability coefficient.

[0011] Preferably, the step of obtaining the multi-source observation time series consistency coefficient is as follows: Obtain the timestamp sequence from the original grid detection data corresponding to each detection sensor within the current inspection cycle; select the timestamp sequence with the largest number of frames from the timestamp sequences corresponding to each detection sensor as the reference time axis; the sensor corresponding to the reference time axis is the reference sensor; for each sensor other than the sensor corresponding to the reference time axis, using each timestamp in the reference time axis as a benchmark, determine the nearest neighbor timestamp with the smallest absolute value of the time difference with each timestamp in the reference time axis in the timestamp sequence, and record it as the minimum nearest neighbor timestamp; and then assign each minimum nearest neighbor timestamp... The absolute difference between the sensor and the timestamp in the corresponding reference time axis is calculated to obtain the time misalignment amplitude of the sensor at the i-th reference timestamp. The time interval sequence between adjacent timestamps is calculated based on the reference time axis, and the median of the time interval sequence is used as the reference time interval. The median of all time misalignment amplitudes of any sensor is taken to obtain the typical time misalignment amplitude of the sensor. The time series consistency score is calculated based on the typical time misalignment amplitude and the reference time interval. The time series consistency scores corresponding to all non-reference sensors are used to form a time series consistency score set, and the minimum value in the time series consistency score set is taken as the multi-source observation time series consistency coefficient.

[0012] Preferably, the steps for obtaining the model operation and maintenance consistency coefficient are as follows: Obtain the set of operation and maintenance event records within the current inspection cycle and the set of model update timestamps for the initial virtual network frame model; for any operation and maintenance event, determine whether there exists a model update timestamp greater than or equal to the event occurrence time; if so, determine that the operation and maintenance event has been synchronized to the initial virtual network frame model; if not, determine that the operation and maintenance event is an unsynchronized operation and maintenance event; obtain the total number of operation and maintenance events within the current inspection cycle, count the total number of unsynchronized operation and maintenance events, and subtract the ratio of the total number of unsynchronized operation and maintenance events to 1 to obtain... The synchronization ratio of operation and maintenance events is calculated as follows: All operation and maintenance events within the current inspection cycle are sorted in ascending order according to their occurrence time to form a time-ordered operation and maintenance event sequence. If two adjacent operation and maintenance events in the time-ordered operation and maintenance event sequence are both determined to be out-of-synchronization events, they are considered to be in the same out-of-synchronization segment. When a state switch occurs between an out-of-synchronization operation and maintenance event and a synchronized operation and maintenance event, it is considered to form a new out-of-synchronization segment. The number of out-of-synchronization segments is counted to obtain the number of consecutive out-of-synchronization segments. The synchronization ratio of operation and maintenance events is divided by the sum of the number of consecutive out-of-synchronization segments and 1 to obtain the model operation and maintenance consistency coefficient.

[0013] Preferably, the step of determining whether the initial virtual space frame model meets the simulation conditions based on the model validity index is as follows: compare the model validity index with the credibility threshold; if the model validity index is greater than or equal to the credibility threshold, then the initial virtual space frame model is determined to meet the simulation conditions; if the model validity index is less than the credibility threshold, then the initial virtual space frame model is determined not to meet the simulation conditions.

[0014] Preferably, the steps for obtaining the corrected virtual space frame model are as follows: Based on the geometric structural state parameters, node coordinates, and structural operation state parameters of each member in the initial virtual space frame model, a one-to-one comparison is performed with the corresponding structural feature data extracted from the original space frame detection data, and the difference between each corresponding parameter is calculated to form a model difference vector; the amplitude of each component in the model difference vector is compared, and the parameter corresponding to the parameter whose difference amplitude is greater than the stable range of the current model parameter change is determined as the parameter to be corrected, and a set of parameters to be corrected is formed; while keeping the spatial topological relationship of the initial virtual space frame model unchanged, a structural constraint relationship between the parameters to be corrected is established based on the member connection relationship and node intersection relationship, so that the continuity of member connection and node geometric constraint conditions are not destroyed during the parameter update process; according to the model difference vector, the model parameters in the set of parameters to be corrected are subjected to iterative correction processing, so that the corrected parameter values ​​gradually approach the structural feature data reflected by the original space frame detection data, while satisfying the structural constraint relationship, and the corrected virtual space frame model is obtained.

[0015] Preferably, a digital twin-based intelligent inspection method for steel structure space frames includes the following steps: Step 1: Automated inspection of the steel structure space frame is performed using detection sensors to collect raw space frame inspection data; Step 2: The raw space frame inspection data is received, and an initial virtual space frame model corresponding one-to-one with the physical entity of the steel structure space frame is constructed based on the raw space frame inspection data; Step 3: Model consistency evaluation data of the initial virtual space frame model is obtained. The model consistency evaluation data includes the verifiability of node connection constraints, the time misalignment amplitude of multiple sensors, and the amount of unsynchronized operation and maintenance events. The model effectiveness index of the virtual space frame model is evaluated based on the model consistency evaluation data, and the model effectiveness index is used to determine whether the initial virtual space frame model meets the simulation conditions; Step 4: If the initial virtual space frame model is determined not to meet the simulation conditions, then... The data assimilation model self-calibration is performed to obtain the calibrated virtual space frame model. The model effectiveness index is obtained again. When the model effectiveness index meets the simulation conditions, the data assimilation model self-calibration is terminated to obtain the final virtual space frame model, which is then transmitted to the evaluation and decision module. If the initial virtual space frame model is determined to meet the simulation conditions, it is directly used as the final virtual space frame model. Step 5: The original space frame detection data is preprocessed, enhanced, and feature extracted to obtain the processed space frame data. Step 6: The processed space frame data is analyzed using a defect identification algorithm model to identify defect information in the steel structure space frame. Step 7: Based on the identified defect information, the defect information is mapped to the final virtual space frame model, mechanical simulation and safety assessment are performed, and maintenance decision suggestions are generated.

[0016] The technical effects and advantages of this invention are as follows: The system acquires model consistency assessment data for the initial virtual space frame model, evaluates the model effectiveness index of the virtual space frame model based on the model consistency assessment data, determines whether the initial virtual space frame model meets the simulation conditions based on the model effectiveness index, and initiates data assimilation-based model self-correction to obtain a corrected virtual space frame model. The model effectiveness index is acquired again, and the data assimilation-based model self-correction is terminated when the model effectiveness index meets the simulation conditions, resulting in the final virtual space frame model. The final virtual space frame model is then transmitted to the evaluation and decision module, effectively reducing the uncertainty in the structural safety management and maintenance decision-making process. Attached Figure Description

[0017] Figure 1 This is a structural diagram of a digital twin-based intelligent detection system for steel space frames, provided in an embodiment of this application.

[0018] Figure 2 A flowchart illustrating an intelligent detection method for steel structure space frames based on digital twins, provided for an embodiment of this application. Detailed Implementation

[0019] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings. In addition, the forms of the various structures described in the following embodiments are merely illustrative. The intelligent detection system and method for steel structure space frame based on digital twins involved in the present invention are not limited to the structures described in the following embodiments. All other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] This invention provides an intelligent inspection system for steel structure space frames based on digital twins, such as... Figure 1 As shown, the system includes: The defect identification module is used to perform automated inspection of steel structure space frame through detection sensors and collect raw space frame inspection data. It should be noted that the detection sensors include one or more of the following: high-definition camera, infrared thermal imager, laser scanner or multispectral camera. The virtual mapping module is used to receive the original space frame inspection data and construct an initial virtual space frame model that corresponds one-to-one with the physical entity of the steel structure space frame based on the original space frame inspection data. The virtual space frame model is used to characterize the geometric state, spatial topology, and structural operation parameters of the steel structure space frame. In this embodiment, it should be specifically explained that the steps for obtaining the initial virtual network structure model are as follows: Receive the raw space frame inspection data collected by the defect identification module, and organize the raw space frame inspection data according to the collection time, spatial location and data type to form a raw inspection data set; Based on the original detection data set, structural feature information such as the outline of the members, the position of the nodes and the boundary of the components in the steel structure space frame is extracted to obtain structural feature data that describes the geometric shape and connection relationship of the space frame; Based on structural feature data, the connection relationships between each member and node in the steel structure space frame are determined, and a space frame spatial topology that reflects the member connection method and node intersection relationship is constructed to characterize the spatial topological relationship of the steel structure space frame. Based on structural feature data and spatial topology, geometric structural state parameters such as spatial position, length direction and node coordinates of each member in the steel structure space frame are determined to characterize the geometric structural state of the steel structure space frame. Based on the construction of geometric structural state parameters and spatial topological relationships, the state parameters related to the operation state of the steel structure space frame are associated with the corresponding rods or nodes and assigned values ​​to form an initial virtual space frame model that includes geometric structural state, spatial topological relationships and structural operation state parameters.

[0021] The credibility assessment module is used to obtain model consistency assessment data of the initial virtual grid model. The model consistency assessment data includes the set of all nodes in the model, the timestamp sequence of each sensor, the set of operation and maintenance event records, and the set of model update timestamps. The model validity index of the virtual grid model is evaluated based on the model consistency assessment data, and the model validity index is used to determine whether the initial virtual grid model meets the simulation conditions. In this embodiment, it should be specifically explained that the steps for obtaining the model efficiency index are as follows: Based on the spatial topology of the initial virtual network model, the set of all nodes in the virtual network model is obtained, and the verifiable coefficient of structural connection is evaluated based on the set of all nodes. Obtain the timestamp sequence from the original grid detection data corresponding to each detection sensor within the current inspection cycle, and evaluate the multi-source observation time series consistency coefficient based on each timestamp sequence. Obtain the set of operation and maintenance event records and the set of model update timestamps of the initial virtual network frame model within the current inspection cycle, and evaluate the model operation and maintenance consistency coefficient based on the set of operation and maintenance event records and the set of model update timestamps. The structural connectivity verifiable coefficients, multi-source observation time-series consistency coefficients, and model operation and maintenance consistency coefficients are normalized. Specifically, in this embodiment, vector normalization can be used to normalize these coefficients. Specifically, the structural connectivity verifiable coefficients, multi-source observation time-series consistency coefficients, and model operation and maintenance consistency coefficients are combined to form a three-dimensional vector. The norm value is obtained by calculating the square root of the sum of squares of each component of this three-dimensional vector, and each coefficient is divided by the norm value to achieve normalization at a unified scale. This normalization method avoids the unbalanced impact on the subsequent calculation of the model effectiveness index caused by differences in the value range, dimensions, or fluctuation amplitude of different coefficients, thereby improving the stability and consistency of the model effectiveness evaluation results. Since vector normalization is a prior art in this field, this embodiment does not further limit its specific calculation process. The model effectiveness index is calculated based on the normalized structural connectivity verifiable coefficients, multi-source observation time-series consistency coefficients, and model operation and maintenance consistency coefficients. The specific steps are as follows: ; In the formula, Represented as the model efficiency index, Represented as the normalized verifiable coefficients of structural connectivity. Represented as the normalized multi-source observation time series consistency coefficient. This is represented as the model operation and maintenance consistency coefficient after normalization. , , This represents the weighted coefficients of the normalized structural connectivity verifiable coefficients, the normalized multi-source observation time series consistency coefficients, and the normalized model operation and maintenance consistency coefficients. , , , Obtained through the analytic hierarchy process, for example , , The weights can be 0.3, 0.6, or 0.1. The Analytic Hierarchy Process (AHP) is an existing technique for determining the weights of multi-factor evaluation indicators. This method constructs a hierarchical structure between the target layer, criterion layer, and indicator layer. It compares the importance relationships between each evaluation factor pairwise to form a judgment matrix, and calculates the weight coefficients corresponding to each evaluation factor based on the judgment matrix, thereby achieving a comprehensive evaluation of multiple indicators. In this embodiment, the weight coefficients are obtained through the AHP and used to calculate the model's effectiveness index. The AHP and its weight determination process are existing technologies in this field, and this embodiment does not further limit its specific calculation steps.

[0022] In this embodiment, it should be specifically explained that the steps for obtaining the verifiable coefficients of the structural connection are as follows: Based on the spatial topology of the initial virtual space frame model, the set of all nodes in the virtual space frame model is obtained, and the set of key nodes is selected from the set of all nodes; wherein, the set of key nodes is the set of nodes that satisfy "at least two members are connected". For each key node in the key node set, the set of connecting members corresponding to the key node is obtained based on the initial virtual space frame model, and the node connection constraints of the key node are determined accordingly; wherein, the node connection constraints include at least: the node coordinates of the key node, the number of members in the set of connecting members of the key node, and the set of included angles between any two members in the set of connecting members. Based on the original grid structure detection data, the observation side node connection feature data corresponding one-to-one with the key node set is extracted; wherein, the observation side node connection feature data includes at least: the coordinates of the observation node corresponding to the key node, the number of observation connecting rods corresponding to the key node, and the set of included angles between the observation connecting rods; For each key node, the node connection constraints of that key node are compared one-to-one with the corresponding observation-side node connection feature data to calculate the node connection constraint residual vector. The node connection constraint residual vector includes at least the Euclidean distance of the node coordinate difference vector, the absolute value of the difference in the number of connecting members, and the maximum absolute value of the difference in the set of included angles. This yields the node connection constraint residual vector used to characterize the degree of deviation between the "model connection relationship" and the "observation connection relationship" of that key node. The norm of the residual vectors of the connection constraints of each key node is calculated based on the residual vectors of the node connection constraints. The verifiability of the node connection of each key node is then calculated based on the norm of the residual vectors of the connection constraints of each key node. The specific steps are as follows: ; In the formula, Let represent the verifiability of node connectivity for the i-th critical node. Let it be represented as the node connection constraint residual vector of the i-th critical node. Let be the norm of the node connection constraint residual vector of the i-th key node; It should be noted that the norm value is used to characterize the overall deviation of the node connection constraint residual vector. The norm value is a fundamental concept in linear algebra, used to measure the magnitude or amplitude of a vector. It can be calculated using Euclidean norm, absolute value norm, or other common vector norm forms. For example, in the Euclidean norm form, the norm value is obtained by summing the squares of each component of the vector and taking the square root. The above-mentioned methods for calculating the norm value are existing techniques in this field, and this embodiment does not limit the specific calculation form of the norm.

[0023] The verifiability of node connections of all key nodes in the key node set is aggregated to obtain the structural connection verifiability coefficient.

[0024] It should be noted that aggregating the verifiability of connections among nodes in the key node set to obtain structural connection verifiability coefficients is a conventional data fusion and comprehensive evaluation technique in this field. Aggregation methods can take various forms, such as geometric mean, arithmetic mean, multiplicative combination, or constructing a unified evaluation function based on the overall residual. All of these different aggregation methods can achieve an overall representation of the verifiability of multiple nodes. This embodiment does not limit the specific aggregation method.

[0025] In this embodiment, it should be specifically explained that the steps for obtaining the multi-source observation time series consistency coefficient are as follows: Obtain the timestamp sequence from the original grid detection data corresponding to each detection sensor within the current inspection cycle. Select the timestamp sequence with the largest number of frames from the timestamp sequences corresponding to each detection sensor as the reference time axis. The sensor corresponding to the reference time axis is the reference sensor. It should be noted that the current inspection cycle refers to the time interval from the start of data collection of the raw grid structure by the detection sensors to the end of data collection during a complete automated inspection. The inspection cycle can be automatically determined by the system based on the start and end times of the inspection task, or it can be divided according to the actual inspection work plan.

[0026] For each sensor other than the sensor corresponding to the reference time axis, using each timestamp in the reference time axis as a reference, determine the nearest neighbor timestamp with the smallest absolute value of the time difference with each timestamp in the timestamp sequence, and denot it as the minimum nearest neighbor timestamp. Then, the absolute difference between each least nearest neighbor timestamp and the corresponding timestamp in the reference time axis is calculated to obtain the time misalignment amplitude of the sensor at the i-th reference timestamp. The time interval sequence between adjacent timestamps is calculated based on a reference time axis, and the median of the time interval sequence is used as the base time interval. For any given sensor, the median of all its time misalignment amplitudes is taken to obtain the typical time misalignment amplitude of that sensor. The timing consistency score is then calculated based on the typical time misalignment amplitude and the reference time interval. The specific steps for obtaining this score are as follows: ; In the formula, Let be the time consistency score of the s-th sensor. This is represented as the typical time misalignment amplitude of the s-th sensor. This is represented as the baseline time interval; The time series consistency scores corresponding to all non-reference sensors are used to form a time series consistency score set, and the minimum value in the time series consistency score set is taken as the multi-source observation time series consistency coefficient.

[0027] In this embodiment, it should be specifically explained that the steps for obtaining the model operation and maintenance consistency coefficient are as follows: Obtain the set of operation and maintenance event records and the set of model update timestamps of the initial virtual network structure model within the current inspection cycle; For any operation and maintenance event, determine whether there is a model update timestamp greater than or equal to the event occurrence time. If there is, it is determined that the operation and maintenance event has been synchronized to the initial virtual network model; if not, it is determined that the operation and maintenance event is an unsynchronized operation and maintenance event. Get the total number of operation and maintenance events in the current inspection cycle, count the total number of unsynchronized operation and maintenance events, and subtract the ratio of the total number of unsynchronized operation and maintenance events to the total number of operation and maintenance events from 1 to get the operation and maintenance event synchronization ratio. All maintenance events within the current inspection cycle are sorted in ascending order according to their occurrence time to form a time-ordered maintenance event sequence. In this time-ordered sequence, if two adjacent maintenance events are both identified as out-of-synchronization events, they are considered to be in the same out-of-synchronization segment. When a state transition occurs between an out-of-synchronization event and a synchronized event, it is considered to form a new out-of-synchronization segment. The number of out-of-synchronization segments is counted to obtain the number of consecutive out-of-synchronization segments. Divide the synchronization ratio of operation and maintenance events by the sum of the number of consecutive unsynchronized segments and 1 to obtain the model operation and maintenance consistency coefficient.

[0028] In this embodiment, it should be specifically explained that the step of determining whether the initial virtual space frame model meets the simulation conditions based on the model effectiveness index is as follows: The model's effectiveness index is compared with a confidence threshold. If the model's effectiveness index is greater than or equal to the confidence threshold, the initial virtual space frame model is determined to meet the simulation conditions; if the model's effectiveness index is less than the confidence threshold, the initial virtual space frame model is determined not to meet the simulation conditions. The confidence threshold is obtained through an adaptive thresholding method, an existing technique for dynamically determining a judgment threshold based on data distribution characteristics. By statistically analyzing the model's effectiveness index sequence within historical calculation periods, its numerical distribution characteristics are obtained, and the confidence threshold for determining whether the model meets the simulation conditions is adaptively determined based on these characteristics. This avoids the incompatibility issues caused by using a fixed threshold under different testing batches or different working conditions. The confidence threshold obtained through the adaptive thresholding method can be adjusted according to the overall changes in the model's effectiveness index, improving the stability and rationality of the judgment results in complex engineering environments. The adaptive thresholding method and its threshold determination process are existing technologies in this field, and this embodiment does not further limit its specific implementation steps.

[0029] If the model self-calibration module determines that the initial virtual space frame model does not meet the simulation conditions, it initiates data assimilation-based model self-calibration to obtain a calibrated virtual space frame model. It then obtains the model effectiveness index again. When the model effectiveness index meets the simulation conditions, it terminates the data assimilation-based model self-calibration to obtain the final virtual space frame model, which is then transmitted to the evaluation and decision module. If the initial virtual space frame model is determined to meet the simulation conditions, it directly uses the initial virtual space frame model as the final virtual space frame model. In this embodiment, it should be specifically explained that the steps for obtaining the corrected virtual space frame model are as follows: Based on the geometric structural state parameters, node coordinates, and structural operation state parameters of each member in the initial virtual space frame model, a one-to-one comparison is made with the corresponding structural feature data extracted from the original space frame detection data. The difference between each corresponding parameter is calculated to form a model difference vector used to characterize the degree of deviation between the model state and the detection data. The magnitudes of each component in the model difference vector are compared, and the parameters whose difference magnitudes are greater than the current model parameter change stability range are identified as parameters to be corrected, forming a set of parameters to be corrected. It should be noted that the current stable range of model parameter changes refers to the natural fluctuation range of each model parameter within a continuous calculation period during historical model updates. This stable range can be obtained through statistical analysis of historical model parameter sequences, for example, by determining the normal fluctuation amplitude range based on the continuous difference sequence of historical model parameters, as a reference range for judging whether the current parameter changes are abnormal. The method for determining the stable range based on the fluctuation characteristics of historical parameters described above is prior art in this field, and this embodiment does not further limit its specific statistical methods.

[0030] While keeping the initial virtual space frame model spatial topology unchanged, structural constraint relationships between the parameters to be corrected are established based on the connection relationships of the members and the intersection relationships of the nodes, so that the continuity of the member connections and the geometric constraints of the nodes are not destroyed during the parameter update process. Based on the model difference vector, iterative correction processing is performed on the model parameters in the set of parameters to be corrected, so that the corrected parameter values ​​gradually approach the structural feature data reflected by the original space frame detection data, while satisfying the structural constraint relationship, thus obtaining the corrected virtual space frame model.

[0031] The data processing module is used to preprocess, enhance, and extract features from the raw space frame inspection data to obtain the processed space frame data. The machine learning recognition module is used to analyze the processed space frame data through a defect recognition algorithm model to identify defect information in the steel structure space frame. The defect information includes the defect type and defect location. It should be noted that the defect recognition algorithm model can include an image defect recognition model based on a convolutional neural network, a defect localization model based on a target detection network, or a defect detection model based on point cloud feature extraction and geometric anomaly analysis. By performing feature extraction, classification, or target localization processing on the detection data, defect information such as cracks, corrosion, and component deformation in the structure can be identified, and the defect type and its corresponding spatial location can be output. The defect recognition algorithm model and the method for analyzing the detection data to obtain defect information are both existing technologies in this field. This embodiment does not provide a detailed description of the specific structure and training process of the defect recognition algorithm model.

[0032] The assessment and decision-making module is used to map the identified defect information to the final virtual space frame model, perform mechanical simulation and safety assessment, and generate maintenance decision recommendations.

[0033] It should be noted that mapping defect information to a virtual space frame model, performing mechanical simulation and safety assessment, and generating maintenance decision recommendations are existing technologies, and this embodiment will not provide a detailed description of their specific steps.

[0034] In this embodiment, it should be specifically explained that, as Figure 2 As shown, a smart inspection method for steel structure space frames based on digital twins includes the following steps: Step 1: Automated inspection of the steel structure space frame is carried out using detection sensors to collect raw space frame inspection data; Step 2: Receive the original space frame inspection data and construct an initial virtual space frame model that corresponds one-to-one with the physical entity of the steel structure space frame based on the original space frame inspection data; Step 3: Obtain the model consistency evaluation data of the initial virtual grid model. The model consistency evaluation data includes the verifiability of node connection constraints, the magnitude of time misalignment of multiple sensors, and the amount of unsynchronized operation and maintenance events. Evaluate the model effectiveness index of the virtual grid model based on the model consistency evaluation data, and determine whether the initial virtual grid model meets the simulation conditions based on the model effectiveness index. Step 4: If the initial virtual space frame model does not meet the simulation conditions, start the data assimilation model self-calibration to obtain the calibrated virtual space frame model. Obtain the model effectiveness index again. When the model effectiveness index meets the simulation conditions, terminate the data assimilation model self-calibration to obtain the final virtual space frame model. Then, transfer the final virtual space frame model to the evaluation and decision module. If the initial virtual space frame model meets the simulation conditions, directly use the initial virtual space frame model as the final virtual space frame model. Step 5: Preprocess, enhance, and extract features from the original space frame inspection data to obtain the processed space frame data; Step 6: Analyze the processed space frame data using a defect identification algorithm model to identify defect information in the steel structure space frame; Step 7: Based on the identified defect information, map the defect information to the final virtual space frame model, perform mechanical simulation and safety assessment, and generate maintenance decision suggestions.

[0035] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

[0036] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A smart inspection system for steel structure space frames based on digital twins, characterized in that, The system includes: The defect identification module is used to automatically inspect the steel structure space frame through detection sensors and collect the original space frame inspection data. The virtual mapping module is used to receive the original space frame inspection data and construct an initial virtual space frame model that corresponds one-to-one with the physical entity of the steel structure space frame based on the original space frame inspection data. The virtual space frame model is used to characterize the geometric state, spatial topology, and structural operation parameters of the steel structure space frame. The credibility assessment module is used to obtain model consistency assessment data of the initial virtual grid model. The model consistency assessment data includes the set of all nodes in the model, the timestamp sequence of each sensor, the set of operation and maintenance event records, and the set of model update timestamps. The model validity index of the virtual grid model is evaluated based on the model consistency assessment data, and the model validity index is used to determine whether the initial virtual grid model meets the simulation conditions. If the model self-calibration module determines that the initial virtual space frame model does not meet the simulation conditions, it initiates data assimilation-based model self-calibration to obtain a calibrated virtual space frame model. It then obtains the model effectiveness index again. When the model effectiveness index meets the simulation conditions, it terminates the data assimilation-based model self-calibration to obtain the final virtual space frame model, which is then transmitted to the evaluation and decision module. If the initial virtual space frame model is determined to meet the simulation conditions, it directly uses the initial virtual space frame model as the final virtual space frame model. The data processing module is used to preprocess, enhance, and extract features from the raw space frame inspection data to obtain the processed space frame data. The machine learning recognition module is used to analyze the processed space frame data through a defect recognition algorithm model to identify defect information in the steel structure space frame; The assessment and decision-making module is used to map the identified defect information to the final virtual space frame model, perform mechanical simulation and safety assessment, and generate maintenance decision recommendations.

2. The intelligent detection system for steel structure space frames based on digital twins according to claim 1, characterized in that: The steps for obtaining the initial virtual space frame model are as follows: Receive the raw space frame inspection data collected by the defect identification module, and organize the raw space frame inspection data according to the collection time, spatial location and data type to form a raw inspection data set; Based on the original detection data set, structural feature information such as the outline of the members, the position of the nodes and the boundary of the components in the steel structure space frame is extracted to obtain structural feature data that describes the geometric shape and connection relationship of the space frame; Based on the structural feature data, the connection relationships between each member and node in the steel structure space frame are determined, and the space frame spatial topology is constructed. Based on structural feature data and spatial topology, the geometric structural state parameters of each member in the steel structure space frame are determined. The geometric structural state parameters include spatial position, length direction and node coordinates. Based on the construction of geometric structural state parameters and spatial topological relationships, the state parameters related to the operation state of the steel structure space frame are associated with the corresponding rods or nodes and assigned values ​​to form an initial virtual space frame model that includes geometric structural state, spatial topological relationships and structural operation state parameters.

3. The intelligent detection system for steel structure space frames based on digital twins according to claim 1, characterized in that, The steps for obtaining the model's efficiency index are as follows: Based on the spatial topology of the initial virtual network model, the set of all nodes in the virtual network model is obtained, and the verifiable coefficient of structural connection is evaluated based on the set of all nodes. Obtain the timestamp sequence from the original grid detection data corresponding to each detection sensor within the current inspection cycle, and evaluate the multi-source observation time series consistency coefficient based on each timestamp sequence. Obtain the set of operation and maintenance event records and the set of model update timestamps of the initial virtual network frame model within the current inspection cycle, and evaluate the model operation and maintenance consistency coefficient based on the set of operation and maintenance event records and the set of model update timestamps. The structural connectivity verifiable coefficient, multi-source observation time series consistency coefficient, and model operation and maintenance consistency coefficient are normalized. The model effectiveness index is then calculated based on the normalized structural connectivity verifiable coefficient, multi-source observation time series consistency coefficient, and model operation and maintenance consistency coefficient.

4. The intelligent detection system for steel structure space frames based on digital twins according to claim 3, characterized in that, The steps for obtaining the verifiable coefficients of the structural connection are as follows: Based on the spatial topology of the initial virtual network model, the set of all nodes in the virtual network model is obtained, and the set of key nodes is selected from the set of all nodes. For each key node in the set of key nodes, obtain the set of connecting members corresponding to the key node based on the initial virtual space frame model, and determine the node connection constraints of the key node accordingly. Based on the extraction of observation-side node connection feature data corresponding one-to-one with the key node set from the original grid structure detection data; For each key node, the node connection constraints of the key node are compared one-to-one with the corresponding node connection feature data of the observation side to calculate the node connection constraint residual vector. The norm of the residual vector of each key node connection constraint is calculated based on the residual vector of the node connection constraint, and the verifiability of the node connection of each key node is calculated based on the norm of the residual vector of the node connection constraint. The verifiability of node connections of all key nodes in the key node set is aggregated to obtain the structural connection verifiability coefficient.

5. The intelligent detection system for steel structure space frame based on digital twin according to claim 3, characterized in that: The steps for obtaining the multi-source observation time series consistency coefficient are as follows: Obtain the timestamp sequence from the original grid detection data corresponding to each detection sensor within the current inspection cycle. Select the timestamp sequence with the largest number of frames from the timestamp sequences corresponding to each detection sensor as the reference time axis. The sensor corresponding to the reference time axis is the reference sensor. For each sensor other than the sensor corresponding to the reference time axis, using each timestamp in the reference time axis as a reference, determine the nearest neighbor timestamp with the smallest absolute value of the time difference with each timestamp in the timestamp sequence, and denot it as the minimum nearest neighbor timestamp. Then, the absolute difference between each least nearest neighbor timestamp and the corresponding timestamp in the reference time axis is calculated to obtain the time misalignment amplitude of the sensor at the i-th reference timestamp. The time interval sequence between adjacent timestamps is calculated based on a reference time axis, and the median of the time interval sequence is used as the base time interval. For any sensor, take the median of all its time misalignment amplitudes to obtain the typical time misalignment amplitude of the sensor, and calculate the timing consistency score based on the typical time misalignment amplitude and the reference time interval. The time series consistency scores corresponding to all non-reference sensors are used to form a time series consistency score set, and the minimum value in the time series consistency score set is taken as the multi-source observation time series consistency coefficient.

6. The intelligent detection system for steel structure space frame based on digital twin according to claim 3, characterized in that: The steps for obtaining the model operation and maintenance consistency coefficient are as follows: Obtain the set of operation and maintenance event records and the set of model update timestamps of the initial virtual network structure model within the current inspection cycle; For any operation and maintenance event, determine whether there is a model update timestamp greater than or equal to the event occurrence time. If so, determine that the operation and maintenance event has been synchronized to the initial virtual network model. If it does not exist, the operation and maintenance event is determined to be an unsynchronized operation and maintenance event; Get the total number of operation and maintenance events in the current inspection cycle, count the total number of unsynchronized operation and maintenance events, and subtract the ratio of the total number of unsynchronized operation and maintenance events to the total number of operation and maintenance events from 1 to get the operation and maintenance event synchronization ratio. All maintenance events within the current inspection cycle are sorted in ascending order according to their occurrence time to form a time-ordered maintenance event sequence. In this time-ordered sequence, if two adjacent maintenance events are both identified as out-of-synchronization events, they are considered to be in the same out-of-synchronization segment. When a state transition occurs between an out-of-synchronization event and a synchronized event, it is considered to form a new out-of-synchronization segment. The number of out-of-synchronization segments is counted to obtain the number of consecutive out-of-synchronization segments. Divide the synchronization ratio of operation and maintenance events by the sum of the number of consecutive unsynchronized segments and 1 to obtain the model operation and maintenance consistency coefficient.

7. The intelligent detection system for steel structure space frame based on digital twin according to claim 1, characterized in that: The steps for determining whether the initial virtual space frame model meets the simulation conditions based on the model effectiveness index are as follows: The model validity index is compared with the confidence threshold. If the model validity index is greater than or equal to the confidence threshold, the initial virtual space frame model is determined to meet the simulation conditions. If the model validity index is less than the confidence threshold, the initial virtual space frame model is determined to not meet the simulation conditions.

8. The intelligent detection system for steel structure space frame based on digital twin according to claim 1, characterized in that: The steps for obtaining the corrected virtual space frame model are as follows: Based on the geometric structural state parameters, node coordinates, and structural operation state parameters of each member in the initial virtual space frame model, a one-to-one comparison is made with the corresponding structural feature data extracted from the original space frame detection data, and the difference between each corresponding parameter is calculated to form a model difference vector. The magnitudes of each component in the model difference vector are compared, and the parameters whose difference magnitudes are greater than the current model parameter change stability range are identified as parameters to be corrected, forming a set of parameters to be corrected. While keeping the initial virtual space frame model spatial topology unchanged, structural constraint relationships between the parameters to be corrected are established based on the connection relationships of the members and the intersection relationships of the nodes, so that the continuity of the member connections and the geometric constraints of the nodes are not destroyed during the parameter update process. Based on the model difference vector, iterative correction processing is performed on the model parameters in the set of parameters to be corrected, so that the corrected parameter values ​​gradually approach the structural feature data reflected by the original space frame detection data, while satisfying the structural constraint relationship, thus obtaining the corrected virtual space frame model.

9. A method for intelligent inspection of steel structure space frames based on digital twins, used to implement the intelligent inspection system for steel structure space frames based on digital twins as described in any one of claims 1-8, characterized in that: Includes the following steps: Step 1: Automated inspection of the steel structure space frame is carried out using detection sensors to collect raw space frame inspection data; Step 2: Receive the original space frame inspection data and construct an initial virtual space frame model that corresponds one-to-one with the physical entity of the steel structure space frame based on the original space frame inspection data; Step 3: Obtain the model consistency evaluation data of the initial virtual grid model. The model consistency evaluation data includes the verifiability of node connection constraints, the magnitude of time misalignment of multiple sensors, and the amount of unsynchronized operation and maintenance events. Evaluate the model effectiveness index of the virtual grid model based on the model consistency evaluation data, and determine whether the initial virtual grid model meets the simulation conditions based on the model effectiveness index. Step 4: If the initial virtual space frame model does not meet the simulation conditions, start the data assimilation model self-calibration to obtain the calibrated virtual space frame model. Obtain the model effectiveness index again. When the model effectiveness index meets the simulation conditions, terminate the data assimilation model self-calibration to obtain the final virtual space frame model. Then, transfer the final virtual space frame model to the evaluation and decision module. If the initial virtual space frame model meets the simulation conditions, directly use the initial virtual space frame model as the final virtual space frame model. Step 5: Preprocess, enhance, and extract features from the original space frame inspection data to obtain the processed space frame data; Step 6: Analyze the processed space frame data using a defect identification algorithm model to identify defect information in the steel structure space frame; Step 7: Based on the identified defect information, map the defect information to the final virtual space frame model, perform mechanical simulation and safety assessment, and generate maintenance decision suggestions.

Images