A fatigue crack detection system for steel structures based on digital twins

Abstract

This invention relates to a fatigue crack detection system for steel structures based on digital twins, comprising: a sensing module for real-time acquisition of fatigue crack monitoring data of the steel structure; a digital twin backend data fusion and management module for parsing and standardizing the fatigue crack monitoring data, constructing a digital twin model of the steel structure to map the physical state of the steel structure and the spatial distribution of fatigue cracks; an analysis module for detecting fatigue cracks in the steel structure based on the fatigue crack monitoring data, updating the detection results to the digital twin model for state synchronization, and providing crack risk warnings; and a visualization module for displaying the fatigue crack monitoring data and the detection and warning results output by the analysis module. Compared with existing technologies, this invention, through the coordinated operation of sensing, fusion, analysis, and visualization, achieves real-time detection, trend prediction, and multi-level early warning of fatigue cracks in steel structures, significantly improving the safety and scientific decision-making during the operation and maintenance phase.

Description

Technical Field

[0001] This invention relates to the field of fatigue crack detection technology for steel structures, and in particular to a fatigue crack detection system for steel structures based on digital twins. Background Technology

[0002] Steel structures are widely used due to their light weight and strong seismic resistance. However, under high-frequency, continuous, and repetitive dynamic loads, steel will experience a decrease in strength and the development of microcracks. These cracks will gradually develop with repeated load application, potentially leading to component tearing. Furthermore, fatigue failure in steel structures is not a type of plastic failure; therefore, it is crucial to detect fatigue cracks promptly and conduct continuous monitoring.

[0003] Traditional fatigue crack detection in steel structures mainly relies on manual inspection. This method suffers from problems such as low detection frequency, insufficient efficiency, and strong subjectivity. Furthermore, manual inspection is difficult to conduct monitoring tasks at multiple points. In addition, under conditions of insufficient lighting, high reflectivity, or complex environments, the accuracy and stability of manual and traditional imaging detection methods are insufficient to meet the requirements for early crack identification, leading to the easy omission of some fatigue cracks in steel structures at an early stage.

[0004] Existing crack detection systems mostly analyze data from single sources, lacking a unified processing and fusion mechanism for multi-source heterogeneous data. During steel structure operation and maintenance, data collected by different detection devices are often scattered across independent storage media, making it difficult to form a complete long-term status archive. For example, manual inspection results may only exist in the form of paper documents or isolated electronic storage, making it difficult to analyze in conjunction with sensor-collected data. This limits the scientific rigor of crack development trend assessment and maintenance decisions.

[0005] The development of Building Information Modeling (BIM), deep learning object detection algorithms, real-time image processing technology, and sensor networks has made it possible to efficiently monitor and assess the condition of fatigue cracks in steel structures. For example, BIM can accurately represent the geometric and mechanical properties of steel structures, intelligent detection algorithms can improve the accuracy and robustness of crack identification, and sensor networks can achieve real-time data acquisition from multiple points. However, in engineering applications, these technologies are often characterized by decentralized deployment and difficulty in interoperability, lacking a systematic architecture that unifies and integrates sensing, data transmission, analysis and processing, and visualization management. This makes the early detection, precise location, and multi-dimensional data collaborative management of fatigue cracks in steel structures still a significant challenge. Summary of the Invention

[0006] The purpose of this invention is to overcome the shortcomings of the existing technology, which lacks a systematic architecture that integrates sensing, data transmission, analysis and processing, and visualization management, and to provide a steel structure fatigue crack detection system based on digital twins.

[0007] The objective of this invention can be achieved through the following technical solutions:

[0008] A fatigue crack detection system for steel structures based on digital twins, comprising:

[0009] The sensing module is used to collect fatigue crack monitoring data of steel structures in real time.

[0010] The digital twin backend data fusion and management module is used to analyze and standardize the fatigue crack monitoring data of the sensing module, and construct a digital twin model of the steel structure to map the physical state of the steel structure and the spatial distribution of fatigue cracks.

[0011] The analysis module, connected to the digital twin backend data fusion and management module, is used to detect fatigue cracks in steel structures based on fatigue crack monitoring data, update the detection results to the digital twin model for status synchronization, and provide crack risk warnings.

[0012] The visualization module is used to display the fatigue crack monitoring data collected by the sensing module, as well as the detection results and crack risk warning results output by the analysis module.

[0013] Furthermore, the fatigue crack monitoring data includes crack monitoring images.

[0014] Furthermore, the analysis module is equipped with a crack location component for extracting crack locations, and the processing procedure of the crack location component includes:

[0015] Feature enhancement preprocessing is performed on crack monitoring images, which includes using multi-directional convolution kernels. The edge responses of crack monitoring images in multiple directions are obtained, and the largest edge response is selected for fusion to obtain the enhanced image. The corresponding calculation expression is:

[0016]

[0017]

[0018] In the formula, To enhance the graph, For direction Edge response, The grayscale value of the crack monitoring image;

[0019] The enhanced image is used to obtain crack detection results through a pre-trained crack detection model. This includes the two-dimensional coordinates of the crack center point and the bounding box size; the crack detection results are inversely calculated according to the image size, and the scaling factor is calculated based on the shooting distance and camera intrinsic parameters to adjust the bounding box size, thus obtaining the crack location result in the actual area.

[0020] Furthermore, the analysis module includes a crack analysis component, the processing steps of which include:

[0021] The binary image A of the crack in the detection result is subjected to dilation and erosion operations using a circular structuring element S. The outer bounding line is obtained by dilation and subtraction. ,in, For expansion operations; the inner bounding line is obtained through erosion operations and taking the difference: ,in, This is a corrosion operation;

[0022] The outer and inner bounding lines are derived into a set of crack contour coordinates.

[0023] Furthermore, the processing procedure of the crack analysis component also includes:

[0024] The complexity coefficient is calculated based on the crack length and contour in the detection results. This complexity coefficient includes the tortuosity coefficient or fractal characteristic value.

[0025] The tortuosity coefficient is the ratio of the crack length to the straight-line distance between the two ends of the crack;

[0026] The fractal characteristic value is the relationship between the number of grids required to cover the crack and the grid size, which is determined by mapping the crack profile onto grids of different sizes.

[0027] Furthermore, the analysis module also includes a crack history evolution analysis component, the processing steps of which include:

[0028] The detection results of fatigue cracks are matched with an existing crack database. The matching methods include spatial neighborhood matching and morphological similarity matching. Spatial neighborhood matching includes calculating the spatial distance between the crack's three-dimensional spatial coordinates in the detection results and cracks in the crack database. If the calculated spatial distance is less than a preset distance threshold, the match is successful. Morphological similarity matching includes calculating the Hausdorff distance between the crack's three-dimensional spatial coordinates in the detection results and cracks in the crack database, or performing similarity scoring based on feature descriptors, to determine whether the match is successful.

[0029] If the detection result of fatigue cracks matches the cracks in the crack database, the geometric parameters and complexity coefficients corresponding to the current fatigue crack detection result are recorded.

[0030] The crack length, width, and complexity in the crack database are interpolated or fitted over time to output the crack trend curve.

[0031] Furthermore, the analysis module also includes a crack development severity assessment component, the processing steps of which include:

[0032] Based on the fatigue crack detection results, the crack length, crack width, and crack tip stress, along with the corresponding complexity, are analyzed over time to determine the length change rate. Width change rate Complexity change rate and the rate of change of stress at the crack tip To calculate the comprehensive intensity index This comprehensive intensity index The calculation expression is:

[0033]

[0034] In the formula, These are the weighting coefficients;

[0035] Based on the calculated comprehensive intensity index The severity level is assessed by comparing it with a predefined severity threshold.

[0036] Furthermore, the analysis module also includes a decision-making component, the processing of which includes:

[0037] Based on the severity assessment results of the crack development severity assessment component, corresponding response and control strategies are determined to provide early warning of crack risks.

[0038] Furthermore, the sensing module includes at least one of an image monitoring unit, a displacement field monitoring unit, a stress and strain field monitoring unit, and a temperature field monitoring unit.

[0039] Furthermore, the visualization module is also used to display the corresponding position on the digital twin model of the steel structure in the form of a virtual sensor or crack marker based on the detection results output by the analysis module, and to reflect the changes in the state of the crack and the crack risk warning results through changes in color, brightness or shape.

[0040] Compared with the prior art, the present invention has the following advantages:

[0041] (1) This invention introduces a crack identification algorithm into the analysis module and combines it with contour extraction and precise positioning algorithms to perform stable and accurate crack identification and location analysis on the real-time collected monitoring data, realize the rapid determination and parameterized recording of crack location and geometric size, and update the identification results to the digital twin model in real time, thereby ensuring that the virtual model and the on-site structural state are accurately synchronized, and realizing large-scale and continuous effective monitoring of fatigue cracks.

[0042] (2) The analysis module of this invention, as the core processing unit of the real-time crack detection system for steel structures, is responsible for sequentially completing crack location, feature quantification, historical evolution recording, development severity calculation, and risk classification judgment according to the predetermined data flow from the sensing module and the digital twin backend, and submitting the processing results to the decision-making and report output stage. This module realizes a closed-loop data link from crack identification to risk management. It not only supports accurate spatial positioning and feature analysis of a single detection, but also combines time series data to perform trend analysis and form a multi-dimensional comprehensive evaluation result. This enables maintenance personnel to grasp the health status of the steel structure in the shortest possible time and take timely measures when risks emerge to avoid accidents.

[0043] (3) This invention constructs a digital twin model that is highly synchronized with the physical site and continuously integrates real-time monitoring data. Combined with the safety threshold set by the system, it analyzes the crack development trend and classifies the risk, realizes the output of audible and visual alarms in the risk state or regular reports in the stable stage, thereby improving the accuracy of crack development prediction and significantly enhancing the initiative and safety of fatigue crack risk management.

[0044] (4) This invention performs data parsing and standardization through the multi-source heterogeneous data access component in the digital twin backend data fusion and management module, standardizes and unifies the heterogeneous data collected by different detection devices and sensing terminals, and transmits the above data to the digital twin platform to achieve centralized storage, fusion analysis and long-term tracking, thereby solving the problem of isolated and unshared data from multiple devices, and facilitating integrated maintenance and cross-regional status management in large-scale steel structure projects. Attached Figure Description

[0045] Figure 1 This is a system architecture diagram of a steel structure fatigue crack detection system based on digital twin provided in an embodiment of the present invention;

[0046] Figure 2 This is a schematic diagram of the structure of a sensing module provided in an embodiment of the present invention;

[0047] Figure 3 This is a schematic diagram of the structure of an analysis module provided in an embodiment of the present invention;

[0048] Figure 4 This is a schematic diagram of the structure of a digital twin backend data fusion and management module provided in an embodiment of the present invention;

[0049] Figure 5 This is a schematic diagram of the structure of a visualization module provided in an embodiment of the present invention. Detailed Implementation

[0050] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.

[0051] Therefore, the following detailed description of the embodiments of the invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the invention without inventive effort are within the scope of protection of the invention.

[0052] It should be noted that similar labels and letters in the following figures indicate similar items. Therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures.

[0053] Example 1

[0054] like Figure 1 As shown, this embodiment provides a fatigue crack detection system for steel structures based on digital twins, including:

[0055] The sensing module is used to collect fatigue crack monitoring data of steel structures in real time.

[0056] The digital twin backend data fusion and management module is used to analyze and standardize the fatigue crack monitoring data of the sensing module, and construct a digital twin model of the steel structure to map the physical state of the steel structure and the spatial distribution of fatigue cracks.

[0057] The analysis module connects to the backend data fusion and management module of the digital twin. It is used to detect fatigue cracks in steel structures based on fatigue crack monitoring data, update the detection results to the digital twin model for status synchronization, and provide early warning of crack risks.

[0058] The visualization module is used to display the fatigue crack monitoring data collected by the sensing module, as well as the detection results and crack risk warning results output by the analysis module.

[0059] The following is a detailed description of each module:

[0060] I. Perception Module

[0061] like Figure 2 As shown, the sensing module is used to collect real-time monitoring data on fatigue cracks in in-service steel structures.

[0062] The sensing module, serving as the sensing end of the analysis system of this invention, is a crucial bridge connecting the physical entity of the steel structure with its digital twin in the information world. Its core task is to collect high-fidelity, real-time dynamic information about the complex operating environment of the steel structure. The stable and accurate operation of this module is fundamental to ensuring the effectiveness of subsequent crack analysis and early warning.

[0063] In this embodiment, a crane beam in a steel structure factory building is preferred as the application scenario. This component is prone to cracks near the supports and in the middle of the span, where the structure is complex and the stress is concentrated. This embodiment adopts a fixed sensing deployment method to continuously monitor specific high-risk locations.

[0064] The sensing module consists of a series of sensing elements deployed at key locations on the steel structure. It can be decomposed into several different functional units according to specific application requirements, including but not limited to image monitoring units, displacement field monitoring units, stress-strain field monitoring units, and temperature field monitoring units. In different application environments, some or all units can be activated to achieve comprehensive perception of the steel structure's condition, depending on the monitoring target and resource configuration.

[0065] Specifically, the sensing module can be further decomposed into several units to achieve real-time monitoring of fatigue cracks in steel structures from different dimensions:

[0066] 1.1 Image Monitoring Unit

[0067] This unit is capable of acquiring image information of key locations of the steel structure within a set time interval, and is the main means for this system to intuitively observe the fatigue crack condition of the steel structure.

[0068] In this embodiment, the image monitoring unit includes functional components such as an image acquisition device, a controllable light source, a light adjustment component, and a protective structure.

[0069] The light source is used to provide a stable lighting environment in situations where there is insufficient light or the structure of the protective barrier is obstructing the light.

[0070] The light adjustment component can reduce image overexposure caused by reflections from the component surface;

[0071] The protective structure is designed to prevent damage to the acquisition components from external impacts, dust, or other environmental factors, and features diffuse light reflection characteristics on the inner surface to optimize imaging quality.

[0072] In this embodiment, the sensing module preferably only enables the image monitoring unit for real-time monitoring, while other monitoring units are not used in the current environment; however, in other embodiments or different application scenarios, monitoring units such as displacement, stress, and temperature can be enabled as needed to achieve multi-dimensional monitoring.

[0073] 1.2 Displacement field monitoring unit (optional configuration)

[0074] This unit focuses on monitoring the deformation and displacement changes of key parts of the steel structure, and is an important means of evaluating the overall stability and performance of the structure.

[0075] In other application scenarios, monitoring targets can be selected from locations such as the support area, mid-span, and important connection nodes. Relative displacement is measured and absolute deformation is calculated by long-term deployment of displacement sensors. This data can be used to verify the accuracy of deformation calculations in numerical models and to capture abnormal displacement trends before cracks appear, providing support for early warning systems.

[0076] 1.3 Stress and strain field monitoring unit (optional configuration)

[0077] This unit is used to sense changes in the stress state of critical load-bearing components, serving as an important basis for assessing structural safety reserves. In other applications, stress-strain sensors can be deployed inside or on the surface of the structure to directly measure the stress and deformation responses of critical sections.

[0078] 1.4 Temperature monitoring unit (optional configuration)

[0079] This unit is used to detect temperature changes and distribution on the surface and inside of steel structures, and to identify the impact of ambient temperature on the structure. In other applications, temperature sensors can be deployed at load-bearing areas, joints, or weld areas to identify temperature change factors that may affect crack formation or propagation through temperature distribution detection.

[0080] Data integration and transmission

[0081] In this embodiment, the image coordinate positioning component of the sensing module is used to process the data collected by the image sensor. The monitored image information is fused with other optional monitoring data through the coordinate positioning algorithm based on image information, and then uniformly sent to the digital twin backend data fusion and management module to form an observation based on image coordinate data and integrating other monitoring data.

[0082] In this embodiment, coordinate localization based on image information is achieved using the image coordinate localization component of the perception module, including:

[0083] Geometric and optical calibration of the camera components is performed to obtain the lens's intrinsic parameters (focal length, principal point position, pixel ratio coefficient, etc.), extrinsic parameters (attitude and position), and camera distortion coefficients. The calibration process can be completed by acquiring specific calibration patterns from multiple perspectives and combining corner detection and model solving methods.

[0084] A mathematical mapping is established between the pixel coordinate system, camera coordinate system, and world coordinate system, and precise transformation between different coordinate systems is achieved through matrix transformation. This transformation relationship can take into account optical distortion correction and parameter optimization to reduce positioning errors.

[0085] The image monitoring data is mapped to the world coordinate system through the above transformation relationship. The component outputs the positioning result data to realize the positioning in three-dimensional space, so that the analysis module can further detect cracks at this spatial location and the visualization module can present the location distribution of cracks in the actual structure.

[0086] The logical role of the sensing module in the system of this invention:

[0087] The sensing module plays the role of the information source in the overall technical solution of this invention, and its logical function is as follows:

[0088] The real-time monitoring data of the steel structure it generates is the sole and most crucial input driving the digital twin model in the analysis module to perform "crack updates." It is based on this feedback from the real physical world that the analysis module can continuously update the distribution and size of cracks in the steel structure through its internal algorithms, clearly showing users the development of cracks in the steel structure.

[0089] II. Digital Twin Backend Data Fusion and Management Module

[0090] like Figure 4 As shown, this invention relates to a real-time fatigue crack detection system for steel structures based on digital twins. The main function of this module is to analyze, standardize, fuse, and manage long-term observation data from the sensing module and multi-source heterogeneous access components, and to provide unified data support and interaction interfaces for the digital twin model, analysis module, and visualization module. This module includes:

[0091] 2.1 Multi-source heterogeneous access components

[0092] Multi-source heterogeneous access components are used to expand the data sources of the system. In addition to the sensing module built into the system, it can also access structural health monitoring devices from different manufacturers or of different types, including third-party testing platforms, sensing terminals deployed in specific scenarios, historical monitoring archives, and data output from edge computing nodes.

[0093] This component connects with the data parsing and standardized processes of the digital twin backend through a customized interface protocol, enabling the detection results from heterogeneous hardware and software environments to be unified in terms of format, coordinate system, and timestamps. This ensures that these external data can participate in the fusion analysis and long-term storage together with the real-time data collected by this system.

[0094] This architecture ensures the system's flexible adaptability under cross-regional and multi-device conditions, and supports rapid deployment across different projects, enabling integrated management and collaborative operation and maintenance of fatigue crack monitoring.

[0095] 2.2 Data Parsing and Standardization Components

[0096] This component is used to perform unified processing on the raw observation data input from the sensing module and the multi-source heterogeneous access component, including formatting, coordinate alignment, time synchronization, outlier removal, and necessary noise reduction.

[0097] In this embodiment, the data parsing and standardization component maps all data from different sources to the world coordinate system, achieving spatial consistency. The processed data is then unidirectionally transferred to the data management and storage component for archiving and subsequent retrieval.

[0098] 2.3 Data Management and Storage Components

[0099] This component is used for centralized management and storage of standardized data. Based on a relational database, it mainly includes:

[0100] Structural Model Information Table: Records the model data (beams, columns, etc.) of the steel structure of the crack display platform, including spatial coordinates, dimensions, material parameters, and model source.

[0101] Detection System Information Table: Records the crack detection terminal's number, world coordinate position, angle parameters, and interface information.

[0102] The detection results data table stores the results such as crack detection time, length, outline shape, and location coordinates, and retains historical records.

[0103] The detection system status data table records the operating status of the edge detection terminal, including CPU / GPU usage, memory usage, temperature, and abnormal logs.

[0104] The data tables are linked by primary keys and foreign keys to form a unified data model, supporting conditional retrieval, index management, version control, and access control.

[0105] The custom service interface component allows bidirectional access to the component, providing real-time and historical data support for the digital twin model and analytics module.

[0106] 2.4 Digital Twin Model of Fatigue Cracks in Steel Structures

[0107] This model is built upon fused data provided by the data management and storage components and constructed based on a 3D steel structure model with BIM information attributes. It can accurately map the physical state of the steel structure and the spatial distribution of fatigue cracks in the information world. In this embodiment, the BIM information attributes are formed by combining the geometric model built by SolidWorks with the parameter information of the structural components. The digital twin model maintains bidirectional data interaction with the customized service interface component to update crack location and size in real time. It can also access long-term crack records from the data management and storage components and transmit them to the analysis module through the customized service interface to analyze the historical development of cracks over long periods.

[0108] 2.5 Customized Service Interface Components

[0109] This component serves as the system's multifunctional entry point for invocation and interaction, supporting functions such as data acquisition, model-driven processing, analysis task assignment, and result output.

[0110] In this embodiment, the customized service interface maintains bidirectional communication with the analysis module to call specific analysis algorithms to process data in the database; and maintains unidirectional communication with the visualization module to push analysis results, model rendering data, or crack development information to the front end for display.

[0111] The logical role of the digital twin backend data fusion and management module in this invention:

[0112] In this invention, the digital twin backend data fusion and management module serves as the core data hub, undertaking the tasks of aggregating, classifying, storing, and distributing information across the entire system's information flow. This module can uniformly standardize data from the sensing module and multi-source heterogeneous access components (including high-resolution monitoring images, sensor structural stress data, environmental parameters, etc.), indexing and archiving data according to data type and timestamps to ensure the direct accessibility of data from different sources and formats in subsequent analysis stages. This module also synchronously receives crack detection results, spatial positioning information, geometric feature data, and risk assessment conclusions output by the analysis module, transforming these analysis results into various presentation formats such as images, text, curves, and tables according to visualization requirements, and transmitting them to the visualization module for real-time rendering and interactive display. Simultaneously, this module provides data backtracking and trend analysis interfaces, supporting the querying and comparison of crack development over long periods, providing continuous data support for assessing crack evolution patterns and structural health trends. Its stable and efficient operation is the fundamental guarantee for ensuring that the digital twin model can achieve real-time updates to crack status and accurate historical trend analysis.

[0113] III. Analysis Module

[0114] like Figure 3As shown, the analysis module deploys a steel structure fatigue crack detection algorithm, which is the core component of the real-time steel structure crack detection system. This part performs computer vision and morphological detection on images acquired by the perception module and stored in the digital twin backend data fusion and management system. Specifically, it includes crack identification and extraction, and can perform unified analysis of historical crack evolution and real-time detected cracks. Based on the real-time assessment results, it determines the warning level and presents it in the visualization module. This module includes:

[0115] 3.1 Crack Positioning Components

[0116] The analysis module includes a crack location component for extracting crack locations. The processing steps of this crack location component include:

[0117] Feature enhancement preprocessing is performed on crack monitoring images, which includes using multi-directional convolution kernels. The edge responses of crack monitoring images in multiple directions are acquired, and the largest edge response is selected and fused to obtain an enhanced image.

[0118] The enhanced image is used to obtain crack detection results through a pre-trained crack detection model. This includes the two-dimensional coordinates of the crack center point and the bounding box size; the crack detection results are inversely calculated according to the image size, and the scaling factor is calculated based on the shooting distance and camera intrinsic parameters to adjust the bounding box size, thus obtaining the crack location result in the actual area.

[0119] The details are as follows:

[0120] The crack localization component receives steel structure monitoring images from the digital twin backend data fusion and management module, and runs a deep learning detection algorithm on an embedded computing platform to automatically identify and locate cracks. This component can stably extract crack region features under different shooting angles, image sharpness, and ambient lighting conditions, and output its two-dimensional location coordinates and bounding box dimensions.

[0121] In this embodiment, the crack localization component uses the YOLO series detection model as its basic framework and performs feature enhancement preprocessing targeting the fatigue crack morphology of steel structures. Let the grayscale value of the input image be g(x,y), and then pass through a multi-directional convolution kernel... Obtain edge responses from multiple directions: The enhanced map is obtained by fusing the responses with the maximum response. The processed image, when input into the crack detection model, significantly improves the detection capability for low-contrast, thin cracks. During the inference phase, the detection model outputs normalized center point, width, and height parameters. This component performs inverse coordinate calculation based on the image size: And calculate the scaling factor based on the shooting distance and camera intrinsic parameters. : .

[0122] In this embodiment, to achieve cross-module spatial positioning, the crack positioning component fuses the two-dimensional detection coordinates with the spatial coordinates output by the sensing module. The sensing module pre-calculates the three-dimensional coordinates (X, Y, Z) through camera calibration and extrinsic parameter calculation. This component uses the two-dimensional detection results to index the corresponding spatial points, realizing the three-dimensional location annotation and bounding box mapping of the crack, thereby generating visualized positioning information in the digital twin platform. This fusion step not only ensures the spatial consistency of the detection results but also improves robustness and positioning accuracy under multi-view and multi-light conditions.

[0123] 3.2 Crack Analysis Component

[0124] The analysis module includes a crack analysis component, the processing steps of which include:

[0125] The binary image A of the crack in the detection result is subjected to dilation and erosion operations using a circular structuring element S. The outer bounding line is obtained by dilation and subtraction. ,in, For expansion operations; the inner bounding line is obtained through erosion operations and taking the difference: ,in, This is a corrosion operation;

[0126] The outer and inner bounding lines are derived into a set of crack contour coordinates;

[0127] The complexity coefficient is calculated based on the crack length and contour in the detection results. This complexity coefficient includes the tortuosity coefficient or fractal characteristic value.

[0128] The tortuosity coefficient is the ratio of the crack length to the straight-line distance between the two ends of the crack;

[0129] Fractal eigenvalues ​​are used to map crack profiles onto meshes of different sizes, determining the relationship between the number of meshes required to cover the crack and the mesh size.

[0130] The details are as follows:

[0131] The crack analysis component is used to perform in-depth analysis of the crack location and morphological characteristics output by the crack location component, in order to determine the crack type, development trend, and potential hazard level. This component can combine image processing technology with structural mechanics assessment models to quantitatively extract parameters such as crack length, width, direction, and distribution location, and establish crack evolution curves.

[0132] In this embodiment, to accurately obtain the contour of the connected region of fatigue cracks in the steel structure, the system first performs morphological processing on the binary image A of the crack. A circular structural element S with a radius of 1 pixel is selected, and its origin reflection set is defined as:

[0133] Expansion and corrosion are respectively: , The outer bounding line is obtained by performing a dilation operation and taking the difference: The inner bounding line is obtained by erosion and taking the difference: For cracks with a width of less than 3 pixels or with drastic angle changes, this invention enlarges the connected region proportionally (preferably by a coefficient of 3) before extraction, and then performs copying and difference, avoiding the overlap of the outer and inner bounding lines and improving the stability of the fine crack outline.

[0134] Regarding the sequential derivation of contour coordinates, this invention employs an eight-neighborhood structure clockwise search algorithm. For each connected region, starting from the first non-zero pixel of the first row, neighboring pixels are checked sequentially. If the value is 1, then it is taken as the next point. The coordinates of this point are divided by the magnification factor for normalization and output. At the same time, the pixel is set to zero and the tracking continues until all contour pixels are deleted.

[0135] The crack contour coordinate set obtained by this method has continuity, closure and proportional consistency.

[0136] In terms of morphological feature quantification, this embodiment, in addition to basic parameters such as length, width, and direction, also calculates the complexity coefficient of the crack, including:

[0137] Tortuousness coefficient : Defined as the ratio of the actual length of the crack to the straight-line distance between its two ends. The larger the value, the more tortuous the crack is and the higher the potential risk of stress concentration.

[0138] Fractal eigenvalues ​​of lattice covering method (Optional embodiment): The crack profile is mapped onto a grid of different sizes, and the relationship between the number of grids required to cover the crack and the grid size is recorded. The calculated fractal dimension is used to describe the degree of nonlinear evolution of the crack profile.

[0139] In another alternative embodiment, the component can acquire local stress field data from the digital twin model and extract the maximum principal stress at the crack tip. And calculate the energy release rate. This formula is used to evaluate the propagation potential of cracks under the current stress environment, where a is the crack half-length and E is the elastic modulus of the material.

[0140] During the positioning output phase of this component, the spatial coordinate system and calibration parameters established by the perception module are invoked to map the coordinates of the center point of the two-dimensional bounding box obtained by crack detection to the three-dimensional spatial position, thereby realizing the real-time binding of the crack position with the digital twin model of the steel structure.

[0141] The analysis results can be used as input parameters for subsequent risk assessments, and can support the generation of a crack status visualization interface in the digital twin platform for operation and maintenance personnel to refer to.

[0142] 3.3 Crack History Evolution Analysis Component

[0143] The analysis module also includes a crack history evolution analysis component, whose processing steps include:

[0144] The detection results of fatigue cracks are matched with an existing crack database. The matching methods include spatial neighborhood matching and morphological similarity matching. Spatial neighborhood matching involves calculating the spatial distance between the crack's three-dimensional spatial coordinates in the detection results and the cracks in the crack database. If the calculated spatial distance is less than a preset distance threshold, the match is successful. Morphological similarity matching involves calculating the Hausdorff distance between the crack's three-dimensional spatial coordinates in the detection results and the cracks in the crack database, or performing similarity scoring based on feature descriptors, to determine whether the match is successful.

[0145] If the detection result of fatigue cracks matches the cracks in the crack database, the geometric parameters and complexity coefficients corresponding to the current fatigue crack detection result are recorded.

[0146] The crack length, width, and complexity of the cracks in the crack database are interpolated or fitted over time to output short-term crack trend curves.

[0147] The details are as follows:

[0148] The crack history evolution analysis component is used to archive, compare, and analyze the morphological parameters, complexity parameters, and mechanical field information of the same crack object at different monitoring time points, so as to depict the evolution process of the crack and provide a basis for predicting future development trends.

[0149] In this embodiment, the crack history evolution component first matches crack instances identified in different detection batches with an existing crack database using the unique identification mechanism of the digital twin platform. The strategy includes:

[0150] Spatial Neighborhood Matching: Based on the three-dimensional spatial coordinates output by the analysis component, calculate the spatial distance between the target crack and the crack in the database. If the distance is less than a set threshold, it is determined to be the same crack instance.

[0151] Morphological similarity matching: The matching results are further confirmed by comparing the Hausdorff distance of the crack contour coordinate set or the similarity score based on the feature descriptor.

[0152] For a successfully matched crack instance, this component records its parameters for each time period in the crack evolution data table, including: geometric parameters (length, average width, direction, etc.), complexity parameters (torsional coefficient, whether fractal characteristic values ​​are recorded), mechanical field parameters (if stress field data is available, it is stored), detection timestamp, and sensor / camera number.

[0153] Two strategies can be used for historical evolution analysis:

[0154] Trend curve generation: Interpolate or fit the changes in crack length, width and complexity over time, and output the curve results.

[0155] Period increment calculation: Calculate the rate of change between adjacent detection periods (such as the length growth rate, the tortuosity change rate), which is used for subsequent use by the severity assessment component.

[0156] This component can optionally display crack evolution animations in the digital twin platform interface, allowing operations and maintenance personnel to intuitively view the crack's expansion path and morphological changes over time.

[0157] 3.4 Crack Development Severity Assessment Component

[0158] The analysis module also includes a crack development severity assessment component. The processing steps for this component include:

[0159] Based on the fatigue crack detection results, the crack length, crack width, and crack tip stress, along with the corresponding complexity, are analyzed over time to determine the length change rate. Width change rate Complexity change rate and the rate of change of stress at the crack tip To calculate the comprehensive intensity index ;

[0160] Based on the calculated comprehensive intensity index The severity level is assessed by comparing it with a predefined severity threshold.

[0161] The details are as follows:

[0162] The crack development severity assessment component is used to determine whether a crack is showing a severe development trend in the near future based on historical evolution data and current monitoring results, and outputs an early warning level. In this embodiment, the component calls upon the periodic incremental data provided by the historical evolution component to calculate a comprehensive severity index. : .

[0163] in: : Length change rate (percentage increase per unit time); : Width change rate; : Rate of change of complexity (rate of change of tortuosity coefficient or fractal eigenvalue); : Rate of change of stress at the crack tip (if mechanical field data are available); The weights for each item can be set by the system based on experience or structural importance.

[0164] Evaluation rules can be based on predefined threshold levels, for example:

[0165] Low intensity: The cracks develop slowly and can be inspected routinely.

[0166] Moderate intensity: It is recommended to shorten the testing cycle;

[0167] High intensity: Immediately plan for repair or reinforcement.

[0168] 3.5 Decision-making component

[0169] The analysis module also includes a decision-making component, whose processing steps include:

[0170] Based on the severity assessment results of the crack development severity assessment component, corresponding response and control strategies are determined to provide early warning of crack risks.

[0171] The details are as follows:

[0172] The decision-making component receives the risk level results output by the multi-level safety threshold risk assessment component and formulates corresponding response and control strategies based on the risk level. At low or potential risk levels, the system can generate structural health reports according to a preset cycle. Combining this with trend information provided by the analysis component and the historical evolution component, the system synchronously pushes the visualization results to the user interface for routine monitoring by maintenance personnel. At medium risk levels, the system can adjust the inspection frequency and push real-time crack development trend predictions, alerting on-site personnel to key areas. At high risk levels, the system will immediately trigger audible and visual alarms, control commands, and emergency plan execution, transmitting alarm data, control commands, and historical crack trend curves to the visualization module, achieving synchronized safety response between the on-site and back-end systems.

[0173] This component, through its linkage with modules for crack location, crack analysis, crack history evolution, crack development severity assessment, and risk assessment, achieves a closed-loop management mechanism from crack identification, feature extraction, trend determination to graded treatment, ensuring that the structural safety monitoring system has the ability to respond promptly and provide precise protection.

[0174] Crack-related multimodal report generation module

[0175] In this embodiment, the crack-related multimodal report generation module automatically collects the latest detection and evaluation results from the analysis module and the digital twin backend when a preset time interval is reached or a high-risk state trigger signal is received, and processes and generates a multimodal structured report containing images, text descriptions, trend charts, etc.

[0176] In normal cycle mode, the system triggers data acquisition at fixed time points via a scheduler, obtaining from the digital twin backend: the current crack location image and its 3D spatial coordinates, geometric feature parameters (length, width, and orientation), growth rate and trend curves provided by the historical evolution component, change rate indicators calculated by the severity assessment component, and multi-level risk level determination results. This data is then fed into the multimodal data processing engine to generate textual descriptions and quantitative result summaries, which are embedded in the report template image area of ​​the high-resolution crack image. Simultaneously, trend curves, risk levels, and prediction conclusions are placed in the chart area. The report generator performs adaptive scaling and encoding optimization on the images and automatically processes text layout and formatting, ultimately exporting to PDF, HTML, and other formats suitable for long-term archiving or interactive viewing.

[0177] In high-risk emergency mode, the module will extend to connect to on-site cameras and audio equipment to collect real-time video clips and audio signals. The audio and video files, along with crack images, text reports, and historical trends, will be packaged into a multimodal fusion report, enabling decision-makers to quickly perceive the health status of the steel structure from multiple dimensions, including visual, auditory, and trend data, thereby improving the timeliness and comprehensiveness of early warnings.

[0178] The logical role of the analysis module in this invention:

[0179] The analysis module, as the core processing unit of the real-time crack detection system for steel structures, is responsible for sequentially processing multi-source data provided by the sensing module and the digital twin backend according to a predetermined data flow. This process includes crack location, feature quantification, historical evolution recording, development severity calculation, and risk classification. The results are then submitted to the decision-making and reporting output stage. This module achieves a closed-loop data link from crack identification to risk management. It not only supports accurate spatial location and feature analysis for single detections but also combines time-series data for trend analysis, generating multi-dimensional comprehensive assessment results. This enables maintenance personnel to grasp the health status of the steel structure in the shortest possible time and take timely measures when risks emerge to prevent accidents.

[0180] IV. Visualization Module

[0181] like Figure 5 As shown, the visualization module is used to integrate and display the real-time monitoring data collected by the perception module and the crack simulation analysis results output by the analysis module. It is the core human-computer interaction interface of the analysis system of this invention. As an intuitive "cockpit" connecting the complex back-end system and the front-end engineering management personnel, this module not only realizes data presentation, but its core task is to deeply integrate and contextualize multi-source, heterogeneous, and dynamic data from the perception and analysis stages, transforming abstract data streams and calculation results into situational information that can be directly understood by management personnel and used for rapid decision-making.

[0182] In a specific implementation of this invention, the visualization module constructs and maintains a three-dimensional digital twin model that is highly faithfully mapped to the physical steel structure on-site. This model encompasses structural geometric entities such as crane beams and embeds crack contour features and status data provided by the analysis module. As the foundational carrier for visualization, the model, within a unified, interactive three-dimensional window, achieves collaborative display of various data through multi-layer overlay and information fusion technologies.

[0183] Real-time presentation of information from the sensing module:

[0184] The visualization module receives the crack distribution location, size parameters, and morphological characteristics transmitted back from the sensing module, and displays them in the corresponding component locations of the 3D model as virtual sensors or crack markers. The color, brightness, or shape of the markers can dynamically change according to real-time data, visually reflecting changes in crack status and risk level. For key crack locations, the system provides an interactive information panel; clicking on it allows users to view historical and real-time crack parameter curves for that location (such as crack length-time relationship), providing an intuitive basis for trend analysis.

[0185] Scientific visualization of the analysis module results:

[0186] To reveal the evolutionary characteristics of cracks that are not directly perceptible to the naked eye, the visualization module performs multi-morphological rendering of the updated state of the digital twin model output by the analysis module, including:

[0187] Crack geometry rendering: The spatial outline of cracks is superimposed on the cross-section or surface of the structural model, and key indicators such as crack width and depth are reflected by color mapping or transparency changes.

[0188] Development speed visualization: The crack development speed calculated by the analysis module is presented as a dynamic color band or trajectory animation, enabling managers to intuitively identify changes in the direction and rate of crack expansion.

[0189] Trend animation: By continuously rendering the crack state at different points in time, a playable evolution animation is formed, simulating the entire process of the crack from its initial occurrence to its current state.

[0190] Through the above methods, the visualization module achieves a unified and integrated display of "historical crack data" (trend retrospection) and "real-time structural status" (dynamic monitoring). In an interactive 3D digital twin scenario, engineering managers can simultaneously obtain the spatial distribution, development speed, and status changes of cracks. This not only helps managers quickly identify risk locations and crack development trends but also allows them to select appropriate inspection and reinforcement plans based on the real-time situation, shortening the response time from problem discovery to action. At the same time, the traceability of historical and real-time information ensures the long-term reliability of maintenance, reinforcement, and operation records, enabling structural safety management to shift from passive response to proactive prevention, significantly improving overall operation and maintenance efficiency and safety assurance capabilities.

[0191] The preferred embodiments of the present invention have been described in detail above. It should be understood that those skilled in the art can make numerous modifications and variations based on the concept of the present invention without creative effort. Therefore, all technical solutions that can be obtained by those skilled in the art based on the concept of the present invention through logical analysis, reasoning, or limited experimentation on the basis of existing technology should be within the scope of protection defined by the claims.

Claims

1. A fatigue crack detection system for steel structures based on digital twins, characterized in that, include: The sensing module is used to collect fatigue crack monitoring data of steel structures in real time. The digital twin backend data fusion and management module is used to analyze and standardize the fatigue crack monitoring data of the sensing module, and construct a digital twin model of the steel structure to map the physical state of the steel structure and the spatial distribution of fatigue cracks. The analysis module, connected to the digital twin backend data fusion and management module, is used to detect fatigue cracks in steel structures based on fatigue crack monitoring data, update the detection results to the digital twin model for status synchronization, and provide crack risk warnings. The visualization module is used to display the fatigue crack monitoring data collected by the sensing module, as well as the detection results and crack risk warning results output by the analysis module. The analysis module is equipped with a crack analysis component, the processing procedure of which includes: The complexity coefficient is calculated based on the crack length and contour in the detection results. This complexity coefficient includes the tortuosity coefficient or fractal characteristic value. The tortuosity coefficient is the ratio of the crack length to the straight-line distance between the two ends of the crack; The fractal feature value is the relationship between the number of grids required to cover the crack and the grid size, which is determined by mapping the crack profile onto grids of different sizes. The analysis module also includes a crack development severity assessment component, the processing steps of which include: Based on the fatigue crack detection results, the crack length, crack width, and crack tip stress, along with the corresponding complexity, are analyzed over time to determine the length change rate. Width change rate Complexity change rate and the rate of change of stress at the crack tip To calculate the comprehensive intensity index This comprehensive intensity index The calculation expression is: In the formula, These are the weighting coefficients; Based on the calculated comprehensive intensity index The severity level is assessed by comparing it with a predefined severity threshold.

2. The fatigue crack detection system for steel structures based on digital twins according to claim 1, characterized in that, The fatigue crack monitoring data includes crack monitoring images.

3. The fatigue crack detection system for steel structures based on digital twins according to claim 2, characterized in that, The analysis module is equipped with a crack location component for extracting crack locations. The processing steps of this crack location component include: Feature enhancement preprocessing is performed on crack monitoring images, which includes using multi-directional convolution kernels. The edge responses of crack monitoring images in multiple directions are obtained, and the largest edge response is selected for fusion to obtain the enhanced image. The corresponding calculation expression is: In the formula, To enhance the graph, For direction Edge response, The grayscale value of the crack monitoring image; The enhanced image is used to obtain crack detection results through a pre-trained crack detection model. This includes the two-dimensional coordinates of the crack center point and the bounding box size; the crack detection results are inversely calculated according to the image size, and the scaling factor is calculated based on the shooting distance and camera intrinsic parameters to adjust the bounding box size, thus obtaining the crack location result in the actual area.

4. The fatigue crack detection system for steel structures based on digital twins according to claim 2, characterized in that, The processing steps of the crack analysis component also include: The binary image A of the crack in the detection result is subjected to dilation and erosion operations using a circular structuring element S. The outer bounding line is obtained by dilation and subtraction. ,in, For expansion operations; the inner bounding line is obtained through erosion operations and taking the difference: ,in, This is a corrosion operation; The outer and inner bounding lines are derived into a set of crack contour coordinates.

5. The fatigue crack detection system for steel structures based on digital twins according to claim 1, characterized in that, The analysis module also includes a crack history evolution analysis component, the processing steps of which include: The detection results of fatigue cracks are matched with an existing crack database. The matching methods include spatial neighborhood matching and morphological similarity matching. Spatial neighborhood matching includes calculating the spatial distance between the crack's three-dimensional spatial coordinates in the detection results and cracks in the crack database. If the calculated spatial distance is less than a preset distance threshold, the match is successful. Morphological similarity matching includes calculating the Hausdorff distance between the crack's three-dimensional spatial coordinates in the detection results and cracks in the crack database, or performing similarity scoring based on feature descriptors, to determine whether the match is successful. If the detection result of fatigue cracks matches the cracks in the crack database, the geometric parameters and complexity coefficients corresponding to the current fatigue crack detection result are recorded. The crack length, width, and complexity in the crack database are interpolated or fitted over time to output the crack trend curve.

6. The fatigue crack detection system for steel structures based on digital twins according to claim 1, characterized in that, The analysis module also includes a decision-making component, the processing of which includes: Based on the severity assessment results of the crack development severity assessment component, corresponding response and control strategies are determined to provide early warning of crack risks.

7. The fatigue crack detection system for steel structures based on digital twins according to claim 1, characterized in that, The sensing module includes at least one of an image monitoring unit, a displacement field monitoring unit, a stress and strain field monitoring unit, and a temperature field monitoring unit.

8. The fatigue crack detection system for steel structures based on digital twins according to claim 1, characterized in that, The visualization module is also used to display the corresponding position on the digital twin model of the steel structure in the form of virtual sensors or crack markers based on the detection results output by the analysis module, and to reflect the changes in the state of the cracks and the crack risk warning results through changes in color, brightness or shape.

Images