A steel structure damage identification system based on deep learning and an identification method thereof

Abstract

The application relates to the technical field of steel structure damage identification, and discloses a steel structure damage identification system based on deep learning and an identification method thereof, which comprises a crack acquisition module, a quantity threshold presetting module and a comparison and coping module.In the application, when the crack quantity is greater than or equal to the crack quantity close threshold, it indicates that the degree of approaching the crack quantity threshold is high, at this time, the crack quantity threshold has been approached, and the probability of safety hazards is high, therefore, for this case, the crack quantity is graded first, and then the time interval of damage detection and identification is shortened in proportion to the increase of the level, so that the frequency of damage detection and identification is improved when the crack quantity threshold is approached, so that problems can be found in time, and safety accidents can be avoided.

Description

Technical Field

[0001] This invention relates to the field of steel structure damage identification technology, specifically to a deep learning-based steel structure damage identification system and method. Background Technology

[0002] Steel structures are widely used in many fields of building construction due to their advantages such as high strength, light weight, high plasticity, and convenient construction. However, during use, steel structures may experience a decrease in load-bearing capacity, stiffness, and stability due to problems such as fatigue damage, corrosion, cracks, reduced stiffness, instability failure, and accidental damage, which can lead to safety issues.

[0003] To prevent problems in steel structures that could lead to safety hazards, damage identification is necessary. There are various methods for damage identification, such as applying a magnetic field to the steel structure and using magnetic particles to pinpoint crack locations, suitable for detecting surface and near-surface defects. Thermal imaging technology can be used to detect temperature differences in the structure, thereby identifying potential problems caused by stress or physical changes.

[0004] Chinese patent CN112131781B discloses a steel structure damage detection method based on a fully connected neural network and a transfer rate function. It utilizes ANSYS software to perform transient analysis on a steel frame model, obtaining the nodal accelerations perpendicular to the detection surface. Using MATLAB software, it performs Fourier transforms on each nodal acceleration and divides their spectra to obtain the transfer rate function. Subtracting these transfer rate functions yields the change in the transfer rate function. This change in the transfer rate function is used as the input parameter of the fully connected neural network, and backpropagation is employed to correct the error, thus obtaining the damage index values ​​at each detection location of the steel structure. This damage detection method requires complex calculations and sophisticated equipment. While it yields more accurate damage detection results, it is difficult to apply on construction sites. Summary of the Invention

[0005] In view of the problems existing in the prior art, the purpose of this invention is to provide a steel structure damage identification system and method based on deep learning, so as to detect and identify steel structure damage and avoid the occurrence of safety hazards.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a deep learning-based steel structure damage identification system, comprising: a crack acquisition module, which acquires image data of a single steel structure surface through a visual sensor, and records the number of cracks on the steel structure surface based on the image data, wherein a crack on the steel structure surface where two endpoints intersect is recorded as one, and a crack on the steel structure surface where two non-endpoints intersect is recorded as two; a quantity threshold preset module, which sets a crack quantity threshold on the steel structure surface based on the model, specifications, and surface area of ​​a single steel structure; and a comparison and response module, which compares the number of cracks on the steel structure surface with the crack quantity threshold, and makes different responses based on the comparison results. If the number of cracks is greater than or equal to the crack quantity threshold, the steel structure is repaired or replaced; if the number of cracks is less than the crack quantity threshold, a secondary judgment is performed.

[0007] In some implementations, during the secondary judgment, a threshold is set for the number of cracks to be close to a threshold. When the number of cracks is less than the threshold, the number of cracks on the surface of the steel structure is compared with the threshold, and different responses are made based on the comparison results.

[0008] In some implementations, if the number of cracks is less than the crack number approaching the threshold, it indicates that the number of cracks on the steel structure surface is far from the crack number threshold, and the steel structure continues to be used; if the number of cracks is greater than or equal to the crack number approaching the threshold, it indicates that the number of cracks is close to the crack number threshold. In this case, the number of cracks is classified according to the degree of proximity to the crack number threshold, and the time interval for damage detection and identification is dynamically adjusted for different levels.

[0009] In some implementations, the crack quantity classification divides the range between the crack quantity near the threshold and the crack quantity threshold into three equal parts. Each part represents a range level, which are ranked from smallest to largest as first-level crack quantity, second-level crack quantity, and third-level crack quantity. The crack quantity level is determined based on the number of cracks on the surface of the steel structure.

[0010] In some implementations, during the dynamic adjustment of the damage detection and identification time interval, an initial damage detection time interval is set. For different levels of crack numbers, the damage detection and identification time interval is shortened proportionally in ascending order, with the initial damage detection and identification time interval as a reference.

[0011] In some implementations, based on the influence of identifying two intersecting crack endpoints as a single crack, the number of recorded cracks is less than the actual number of cracks on the steel structure surface. Taking into account the damage detection identification time interval and the conditions of being far from or exceeding the crack number threshold, it is determined that the number of recorded cracks is less than the actual number of cracks, causing the number of first-level cracks to deviate from the level. The level of deviation is then judged three times based on the actual number of cracks to correct the level.

[0012] In some implementations, during the three-stage judgment process, the number of recorded cracks is re-identified in detail to identify the number of intersections between the endpoints of two cracks. The actual number of cracks is the sum of the recorded number of cracks and the number of intersections between the endpoints of two cracks. Based on the total number of cracks, the crack level into which the crack falls is re-determined, and then a new response is taken.

[0013] In some implementations, the criterion for determining whether there is a boundary at the intersection of the endpoints of two cracks is whether there are two obvious segments in the crack forming a distinct angle. This angle is set to 90°. If there are two segments in the crack with an angle of less than 90° between them, it is determined to be two cracks; otherwise, it is a single crack.

[0014] This invention also provides the following technical solutions:

[0015] This invention further provides a deep learning-based method for steel structure damage identification, comprising the following steps: Step 1, acquiring surface image data of the steel structure through a visual sensor, and obtaining the number of surface cracks based on the image data; Step 2, setting a crack number threshold according to the steel structure specifications and surface area; Step 3, comparing the number of cracks with the crack number threshold; if the number of cracks is equal to or greater than the crack number threshold, repairing or replacing the structure; if the number of cracks is less than the crack number threshold, performing a secondary judgment; Step 4, setting a crack number approach threshold; in the secondary judgment, comparing the number of cracks with the crack number approach threshold; if the number of cracks is less than the crack number approach threshold, the steel structure is considered to be in normal use; if the number of cracks is greater than or equal to the crack number approach threshold, first classifying the number of cracks, and then dynamically adjusting the damage detection and identification time interval according to the crack number level.

[0016] The present invention further provides a computer-readable storage medium storing a computer program, which is executed by a processor to implement the above-described deep learning-based steel structure damage identification system.

[0017] The technical solution provided by this invention has the following advantages compared with the prior art:

[0018] Firstly, in this invention, the recorded number of cracks is compared with a crack number threshold. When the number of cracks is greater than or equal to the crack number threshold, it indicates that the number of cracks exceeds the range of normal use and needs to be repaired or replaced. When the number of cracks is less than the crack number threshold, different judgments are made because the degree of less than is different. Specifically, a crack number approach threshold is set, and the number of cracks is compared with the crack number approach threshold. When it is less than the crack number approach threshold, it indicates that the degree of approach to the crack number threshold is low, and the steel structure can be used normally. When it is greater than or equal to the crack number approach threshold, it indicates that the degree of approach to the crack number threshold is high. At this point, it has approached the crack number threshold, and the probability of safety hazards increases. Therefore, in this case, the number of cracks is first classified, and then the time interval of damage detection and identification is shortened proportionally according to the increase of the level, so as to increase the frequency of damage detection and identification when approaching the crack number threshold, so as to detect problems in time and avoid safety accidents.

[0019] Secondly, when recording the number of cracks, for ease of identification, two intersecting cracks are considered as one. This facilitates the initial recording of the number of cracks, but it leads to the problem that the actual number of cracks is greater than the recorded number. This problem of the actual number being greater than the recorded number has little impact on cracks in other intervals, but it mainly affects the interval with the number of first-level cracks. Therefore, for this interval, more detailed identification is added to obtain the actual number of cracks, and then it is determined whether the number of first-level cracks in this interval has deviated from the level. If it deviates, it is corrected to the corresponding level, and then corresponding processing is performed. Attached Figure Description

[0020] Figure 1 A schematic diagram showing the intersection of the endpoints of two cracks;

[0021] Figure 2 This is a schematic diagram of two cracks intersecting at non-endpoints.

[0022] Figure 3 Schematic diagram for identifying the boundary of the crack endpoint;

[0023] Figure 4 This is a schematic diagram of the logical structure of the invention. Detailed Implementation

[0024] 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. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0025] It is understood that the term "a" should be understood as "at least one" or "one or more", that is, in one embodiment, the number of an element can be one, while in another embodiment, the number of the element can be multiple, and the term "a" should not be understood as a limitation on the number.

[0026] The steel structure damage identification system based on deep learning provided by this invention, such as... Figure 4 As shown, it includes:

[0027] Crack Acquisition Module: This module acquires image data of a single steel structure surface using a vision sensor. This image data records the number of cracks on the steel structure surface, which is then used as a basis for subsequent judgment. These surface cracks in steel structures can exhibit various forms and may be intertwined. Therefore, to accurately record the number of surface cracks, a method for determining the number of cracks is presented here. For example... Figure 1 As shown in the diagram, two cracks intersect at the point of intersection, which is the endpoint. In this case, it is considered as a single crack. Figure 2 As shown in the image, two cracks intersect at the middle point, not at their endpoints. This non-endpoint intersection is identified as two cracks. The core function of this module is to accurately capture image data of the steel structure surface using advanced vision sensor technology, especially recording crack information. Through image analysis technology, it identifies the number and shape of cracks, paying particular attention to crack interweaving. Specifically, for interweaving where the intersection point is at the crack endpoint, the system identifies it as a single crack; while for non-endpoint interweaving, it identifies it as two cracks. This precise judgment mechanism ensures accurate recording of the number of cracks, providing reliable data support for subsequent crack assessment and management.

[0028] Crack Quantity Preset Module: This module sets a crack quantity threshold for the surface of a single steel structure based on its model, specifications, and surface area. This threshold represents the maximum acceptable number of cracks on the steel structure's surface. If this threshold is exceeded, the steel structure is deemed unusable and requires repair or replacement. This module is responsible for setting a reasonable crack quantity threshold based on the steel structure's model, specifications, and surface area as an assessment standard for structural safety. The set crack quantity threshold is based on industry standards, specific application environments, and load requirements to ensure a balance between safety and service life. The crack quantity threshold is determined not only based on theoretical calculations but also adjusted in conjunction with field experience and historical data.

[0029] The comparative response module compares the recorded number of cracks on the steel structure surface with a crack count threshold and takes different actions based on the comparison results. If the number of cracks is greater than or equal to the crack count threshold, it indicates that the number of cracks on the steel structure surface has exceeded the minimum acceptable limit, and the steel structure can no longer be used, otherwise it may pose a certain safety hazard. In this case, the steel structure needs to be repaired. If the repair cannot remedy the defect of the steel structure, it needs to be replaced. If the number of cracks is less than the crack count threshold, it indicates that the number of cracks on the steel structure is still within the acceptable range and can continue to be used. In this case, the steel structure is reassessed. This module compares the detected number of cracks with a preset crack count threshold to determine whether the steel structure needs further measures. If the number of cracks is equal to or exceeds the crack count threshold, the structure is considered to have a potential safety risk and needs to be repaired or replaced. When the number of cracks is less than the crack count threshold, the structure is considered to be within the safe range. This comparative analysis mechanism helps managers make quick decisions to ensure the stability and safety of the steel structure. This systematic approach to assessment can optimize maintenance processes, improve work efficiency, and effectively reduce the risk of accidents caused by cracks.

[0030] Secondary Judgment Process: When the number of cracks on the steel structure surface is less than the crack number threshold, there are two possibilities: the degree of less than the crack number threshold may be low or high. If the degree of less than the crack number threshold is low, it means that although the number of cracks on the steel structure surface has not yet reached the crack number threshold, it is close to it. In this case, compared with the case of a high degree of less than the crack number threshold, the steel structure is more likely to have problems. Therefore, in the secondary judgment, a crack number approach threshold is set, which is less than the crack number threshold. The number of cracks on the steel structure surface is compared with the crack number approach threshold, and different actions are taken based on the comparison results. If the number of cracks is less than the crack number approach threshold, it means that the number of cracks on the steel structure surface has not yet approached the crack number threshold, and in this case, the steel structure can continue to be used. If the number of cracks is greater than or equal to the crack number approach threshold, it means that the number of cracks on the steel structure surface has approached the crack number threshold. In this case, the number of cracks on the steel structure surface is classified according to the degree of approach to the crack number threshold, and the time interval of damage detection and identification is dynamically adjusted for different levels of crack numbers.

[0031] The process of classifying the number of surface cracks in steel structures is as follows: The range between the crack number threshold and the near-threshold is divided into three intervals: the first interval, the second interval, and the third interval. The first interval is closer to the near-threshold, and the third interval is closer to the threshold. When the number of surface cracks in the steel structure is less than the threshold but greater than or equal to the near-threshold, the number of cracks at this point is recorded, and the cracks are classified according to the interval they fall into. Cracks falling into the first interval are classified as level one cracks, those falling into the second interval as level two cracks, and those falling into the third interval as level three cracks.

[0032] The process of dynamically adjusting the damage detection and identification time interval is as follows: Initially, the damage detection and identification time interval for the steel structure is set to once every N days. When the number of cracks on the steel structure surface is between a threshold and a certain threshold, the damage detection and identification time interval is shortened proportionally as the number of cracks increases. This is because the likelihood of safety hazards in the steel structure increases with the number of cracks; therefore, shortening the damage detection and identification time interval is crucial for timely problem detection and prevention of safety hazards.

[0033] In summary, in identifying surface crack damage in steel structures, when the number of cracks is below but close to a crack number threshold, a more detailed analysis is required to ensure safety. A "crack number approaching threshold" is set as a reference point. If the number of cracks is higher than or equal to this value, hierarchical management is needed based on its proximity to the crack number threshold. This process subdivides the crack number from the approach threshold to the crack number threshold into three intervals, thereby conducting level one, level two, and level three risk assessments to determine corresponding maintenance measures. Level one risk represents the minimum safety hazard, while level three risk indicates proximity to the crack number threshold, requiring more frequent inspections and potential repairs. Therefore, a dynamic adjustment mechanism is used for the damage detection and identification time interval, initially once every N days, and the inspection interval is proportionally shortened as the crack level increases, ensuring timely detection and handling of potential structural problems. This method not only optimizes resource utilization but also improves the response speed to changes in structural condition, ensuring the safety and reliability of the steel structure throughout its entire lifecycle.

[0034] For example, assuming the threshold for the number of cracks on a single steel structure surface is 100, and the initial condition is that the number of cracks on the steel structure surface obtained by the visual sensor is 70, since 70 is less than 100, a secondary judgment is performed. The threshold for the number of cracks on the steel structure surface is set to be 70, and the range between the near-threshold and the actual threshold is 60-100. This range is divided into three equal parts: 70 (inclusive) - 80, 80 (inclusive) - 90, and 90 (inclusive) - 100. That is, the number of first-level cracks is 70 (inclusive) - 80, the number of second-level cracks is 80 (inclusive) - 90, and the number of third-level cracks is 90 (inclusive) - 100. Since the number of cracks on the steel structure surface is 70, it is a first-level crack count. Since the damage detection and identification time interval under the initial conditions is once every N days, the time interval should be shortened proportionally. For example, it can be shortened to once every 0.9N days, and so on. If it is a second-order crack, the time interval is once every 0.8N days; if it is a third-order crack, the time interval is once every 0.7N days.

[0035] Furthermore, because the intersection of two crack endpoints is considered as one when obtaining the number of cracks on the steel structure surface, the recorded number of cracks may be less than the actual number of cracks on the steel structure surface. However, this problem has different effects on the number of cracks in different intervals. Based on the above, the crack number intervals are divided into five: Interval A, crack number < crack number close to threshold; Interval B, crack number close to threshold ≤ crack number < maximum number of first-order cracks; Interval C, maximum number of first-order cracks ≤ crack number < maximum number of second-order cracks; Interval D, maximum number of second-order cracks ≤ crack number < crack number threshold; Interval E, crack number ≥ crack number threshold. For clarity, based on the above examples, the five intervals are: Interval A, crack number < 70; Interval B, 70 ≤ crack number < 80; Interval C, 80 ≤ crack number < 90; Interval D, 90 ≤ crack number < 100; Interval E, crack number ≥ 100. For interval A, since the number of cracks is far from the crack number threshold, even if two cracks are identified as one, resulting in an actual number of cracks exceeding the recorded number, the impact on this interval is negligible. For interval E, the number of cracks in this interval is already greater than or equal to the crack number threshold, requiring repair or replacement. Therefore, even if two cracks are identified as one, resulting in an actual number of cracks exceeding the recorded number, it has no impact on this interval. For intervals B, C, and D, in summary, the number of cracks is close to the threshold. If two cracks in this comprehensive interval are identified as one, resulting in an actual number of cracks exceeding the recorded number, the final impact could lead to interval B becoming interval C, interval C becoming interval D, and interval D becoming interval E, etc. For interval C, which is already a secondary crack number interval, the damage identification time interval has already been shortened twice. Therefore, if there is a problem with the steel structure, it will not go undetected. Similarly, the same applies to interval D. The detection and identification time intervals for both intervals have been shortened proportionally at least twice. In the worst-case scenario, if no damage is detected during the first damage detection and identification, the next damage detection and identification will not be long after the first one, so the probability of a safety hazard going undetected is low. Therefore, if two cracks are identified as one, leading to an actual number of cracks exceeding the recorded number, the impact is greatest on interval B. In this interval, the damage detection time interval has only been shortened for the first time. If there are many instances of two cracks coinciding as one, and the next damage detection and identification is delayed, it could lead to a safety hazard not being detected in a timely manner. Therefore, for interval B, three additional checks are performed.

[0036] The process involves three stages: When the number of cracks in the steel structure falls within interval B (i.e., the number of first-order cracks), the number of cracks, L, is recorded. Simultaneously, a detailed identification is performed to determine the number of intersections between the endpoints of two cracks on the steel structure surface. Let X be the number of intersections. The actual number of cracks in interval B is then L+X. Next, it is determined which interval L+X falls into, and corresponding adjustments are made accordingly. If 70 ≤ L+X < 80, the cracks remain within interval B. Although the number of cracks has been corrected, it has not yet exceeded the range of interval B. The detection cycle remains at once every 0.9N days to continue observing the crack trend. If 80 ≤ L+X < 90, the cracks enter interval C. After correction, the number of cracks reaches the range of interval C. At this point, the cracks are in the second-order crack state, and the detection cycle is shortened to once every 0.8N days. If 90 ≤ L + X < 100, then the problem enters interval D. After correction, the number of cracks reaches the range of interval D. At this point, the cracks are in the third-order crack state, and the cycle is shortened to once every 0.7N days. If L + X ≥ 100, then the problem enters interval E. After correction, the number of cracks reaches the range of interval E. At this point, the steel structure needs to be repaired or replaced.

[0037] In summary, based on the zoning and identification strategy for the number of cracks in steel structures, the number of cracks is divided into five zones: Zone A is less than 70 cracks, Zone B is 70 (inclusive) to 80, Zone C is 80 (inclusive) to 90, Zone D is 90 (inclusive) to 100, and Zone E is 100 and above. Within this framework, Zone B is the primary focus because during crack identification and detection, two intersecting cracks are treated as one, leading to an actual crack count higher than the recorded value, thus having the greatest impact on Zone B. Therefore, an additional three-step assessment is employed: first, the number of cracks L in Zone B is recorded; then, the number X of intersecting cracks is identified in detail, and the actual crack count is corrected to L+X; finally, the new zone it falls into is determined based on the corrected count. If the corrected count remains within Zone B, the detection cycle remains at 0.9N days; if it enters Zone C, it is considered a Class II crack, and the detection cycle is shortened to 0.8N days; if it enters Zone D, it is considered a Class III crack, and the cycle is reduced to 0.7N days; if it reaches Zone E, immediate repair or replacement is required. This strategy ensures that potential safety hazards can be identified and addressed promptly, even if there is overlap, during the crack identification process.

[0038] It is worth mentioning that the boundary at the endpoints of two cracks may not be obvious in reality. Therefore, the basis for determining whether there is a boundary at the intersection of the endpoints of two cracks is whether there are two distinct segments forming a clear angle within the crack. This angle is set to 90°. That is, if there are two segments within the crack with a clear angle of less than 90°, it is determined to be two cracks; otherwise, it is considered a single crack. Figure 3 As shown in Figure 1, the angles at the two corners are significantly less than 90°, therefore Figure 1 identifies them as two cracks. Figure 3 As shown in Figure 2, the angle at the two corners is significantly greater than 90°, so Figure 2 is identified as a crack.

[0039] The processes described above with reference to the flowcharts in the embodiments disclosed in this invention can be implemented as computer software programs. Embodiments of this invention include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication component, and / or installed from a removable medium. When the computer program is executed by a central processing unit, it performs the functions defined in the methods of this application. It should be noted that the computer-readable medium described above in this application can be a computer-readable signal medium or a computer-readable storage medium, or any combination of the two. A computer-readable storage medium can be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection having one or more wire segments, a portable computer disk, a hard disk, a random access memory, a read-only memory, an erasable programmable read-only memory, an optical fiber, a portable compact disk read-only memory, an optical storage device, a magnetic storage device, or any suitable combination thereof. In this application, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in connection with an instruction execution system, apparatus, or device. In this application, a computer-readable signal medium can include a data signal propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals can take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. A computer-readable signal medium can also be any computer-readable medium other than a computer-readable storage medium, which can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device. The program code contained on the computer-readable medium can be transmitted using any suitable medium, including but not limited to: wireless segments, wire segments, optical cables, RF, etc., or any suitable combination thereof.

[0040] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of the present invention. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, may be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0041] Those skilled in the art should understand that the above description is only a specific embodiment of this application, but the protection scope of this application is not limited thereto. Any changes 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 protection scope of this application.

Claims

1. A deep learning-based steel structure damage identification system, characterized in that, include: The crack acquisition module acquires image data of a single steel structure surface through a vision sensor and records the number of cracks on the steel structure surface based on the image data. Specifically, a crack on the steel structure surface where two endpoints intersect is counted as one, and a crack on the steel structure surface where two non-endpoints intersect is counted as two. The quantity threshold preset module sets the crack quantity threshold on the surface of the steel structure based on the model, specifications, and surface area of ​​a single steel structure. The comparative response module compares the number of cracks on the steel structure surface with a crack quantity threshold and takes different actions based on the comparison result. If the number of cracks is greater than or equal to the crack quantity threshold, the steel structure is repaired or replaced. If the number of cracks is less than the crack quantity threshold, a secondary judgment is performed. In the secondary judgment, a threshold is set for the number of cracks approaching the crack quantity threshold, and a threshold is set for the number of cracks approaching the crack quantity threshold. When the number of cracks is less than the crack quantity threshold, the number of cracks on the steel structure surface is compared with the crack quantity approaching the crack quantity threshold, and different actions are taken based on the comparison result. If the number of cracks is less than the crack quantity approaching the crack quantity threshold, it means that the number of cracks on the steel structure surface is far from the crack quantity threshold, and the steel structure can continue to be used. If the number of cracks is greater than or equal to the crack quantity approaching the crack quantity threshold, it means that the number of cracks is close to the crack quantity threshold. At this time, the number of cracks is classified according to the degree of proximity to the crack quantity threshold, and the time interval of damage detection and identification is dynamically adjusted for different levels. In the crack quantity classification, the range between the crack quantity close to the threshold and the crack quantity threshold is divided into three equal parts. Each part represents an interval level, which are ranked from smallest to largest as first-level crack quantity, second-level crack quantity, and third-level crack quantity. The crack quantity on the surface of the steel structure is used to determine which level of crack quantity it is. In the dynamic adjustment of the damage detection and identification time interval, an initial damage detection time interval is set. For different levels of crack numbers, the damage detection and identification time interval is shortened proportionally in ascending order, with the initial damage detection and identification time interval as a reference. Based on the influence of identifying two intersecting crack endpoints as a single crack, the number of recorded cracks is less than the actual number of cracks on the steel structure surface. Considering the damage detection identification time interval and the conditions of being far from or exceeding the crack number threshold, it is determined that the number of recorded cracks is less than the actual number of cracks, causing the number of first-level cracks to deviate from the level. The level of deviation is then judged three times based on the actual number of cracks to correct the level.

2. The deep learning-based steel structure damage identification system according to claim 1, characterized in that, During the three judgment processes, the number of recorded cracks is re-identified in detail to identify the number of intersections between the endpoints of two cracks. The actual number of cracks is the sum of the number of recorded cracks and the number of intersections between the endpoints of two cracks. Based on the total number of cracks, the crack level of the cracks is re-determined, and then a new response is taken.

3. The deep learning-based steel structure damage identification system according to claim 2, characterized in that, The basis for determining whether there is a boundary at the intersection of the endpoints of two cracks is whether there are two obvious segments in the crack forming a clear angle. This angle is set to 90°. If there are two segments in the crack with a clear angle of less than 90° between them, it is determined to be two cracks; otherwise, it is a single crack.

4. A deep learning-based method for steel structure damage identification, used to implement the system described in any one of claims 1-3, characterized in that, Includes the following steps: Step 1: Acquire surface image data of the steel structure using a visual sensor, and determine the number of surface cracks based on the image data; Step 2: Set a threshold for the number of cracks based on the specifications and surface area of ​​the steel structure; Step 3: Compare the number of cracks with the crack number threshold. If the number of cracks is equal to or greater than the crack number threshold, repair or replace the cracks. If the number of cracks is less than the crack number threshold, perform a second judgment. Step 4: Set the crack count close to the threshold. In the secondary judgment, compare the crack count with the crack count close to the threshold. If the crack count is less than the crack count close to the threshold, the steel structure can be used normally. If the number of cracks is greater than or equal to the threshold, the number of cracks is first classified, and then the damage detection and identification time interval is dynamically adjusted according to the level of the number of cracks.

5. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that is executed by a processor to implement a deep learning-based steel structure damage identification system as described in any one of claims 1-3.

Images