Method for detecting quality of fabricated steel structural member based on image data

Abstract

The application discloses an image data-based quality detection method for fabricated steel structure components, and relates to the technical field of steel structure welding quality detection.The method can effectively identify microscopic defects (such as embrittlement phenomenon) through detailed image analysis of the heat-affected zone, and has higher sensitivity and accuracy than traditional detection techniques, which helps to ensure the quality of the welded joints.The identification and evaluation of embrittlement can discover potential structural risks in advance, reduce the probability of structural failure, and improve the overall structural safety by accurately marking the embrittlement area and taking targeted measures.The calculation of the stable area proportion value Pwd and the embrittlement area proportion value Pfx can realize mutual verification of data, and if the results are similar, it indicates that the data is accurate and reliable.This double verification mechanism comprehensively reflects the quality state of the components, ensures the effective screening of qualified products, and provides an optimized basis for subsequent welding quality control and detection strategies through a dynamic feedback mechanism.

Description

Technical Field

[0001] This invention relates to the field of steel structure welding quality inspection technology, specifically a method for quality inspection of prefabricated steel structure components based on image data. Background Technology

[0002] With the increasing demands for construction efficiency and structural safety in the construction industry, prefabricated steel structures, as an emerging building form, have been widely used in modern engineering. This type of structure, through the factory production of prefabricated components, can significantly reduce on-site construction time and labor costs. However, the key to prefabricated steel structures lies in the welding quality of their joints, especially the heat-affected zone, which often becomes a weak point in structural performance.

[0003] During welding, the concentrated heat input alters the microstructure of the metal in the weld zone, leading to increased inhomogeneity and brittleness in the material's mechanical properties. This can potentially cause structural failure during later use. Therefore, quality control of welded joints requires more than just ensuring the appearance of the joint; it necessitates a thorough analysis of the welding process's impact on the material's microstructure.

[0004] Traditional inspection methods, such as visual inspection, ultrasonic testing, and X-ray imaging, can often only identify macroscopic defects (such as cracks and pores) and lack sensitivity to microscopic defects (such as embrittlement of the heat-affected zone). In particular, embrittled areas may cause further damage to the heat-affected zone during subsequent processing or re-welding, leading to greater quality problems. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a method for quality inspection of prefabricated steel structure components based on image data, thereby solving the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a method for quality inspection of prefabricated steel structure components based on image data, comprising the following steps:

[0007] S1. Size Classification: Assign a unique identification number to each prefabricated steel structure component to be inspected after welding is completed, and classify it by size after identifying size characteristics to obtain several size groups;

[0008] S2. Image Recognition: Collect microscopic image data of each weld joint of the prefabricated steel structure components to be inspected after welding in several size groups. Each weld joint represents a heat-affected zone. Extract the heat-affected zone features from the microscopic image data. When identifying the heat-affected zone of the steel structure component each time, the heat-affected zone is labeled and the corresponding microscopic image is obtained. Perform feature recognition on the microscopic images according to the feature extraction order of the heat-affected zone. After completing the recognition of the features of each heat-affected zone, establish a feature dataset and obtain: the first embrittlement recognition value qx1, the second embrittlement recognition value qx2, and the third embrittlement recognition value qx3.

[0009] S3. Embrittlement determination: Based on the first embrittlement identification value qx1, the second embrittlement identification value qx2 and the third embrittlement identification value qx3, an embrittlement risk coefficient Rch is constructed. Based on the comparison between the embrittlement risk coefficient Rch and the embrittlement threshold, the relative influence of each embrittlement phenomenon is determined, and the embrittled area and the non-embrittled area are marked.

[0010] S4. Dual verification judgment: Statistically calculate the stable area ratio value Pwd and the embrittled area ratio value Pfx, and evaluate whether each component to be tested is a qualified product based on the stable area ratio value Pwd and the embrittled area ratio value Pfx.

[0011] Preferably, grouping by several sizes includes:

[0012] The surface area of ​​the prefabricated steel structure component to be inspected after welding is less than 1m². 3 The components to be tested are classified into the first small component group;

[0013] The surface area of ​​the prefabricated steel structure component to be inspected after welding is equal to 1m². 3 And less than 10m 3 The components to be tested are classified into the second medium-sized component group;

[0014] The surface area of ​​the prefabricated steel structure components to be inspected after welding is greater than 10m². 3 The components to be tested are classified into the third large component group.

[0015] Preferably, a feature dataset is obtained based on microscopic images to identify several features of the heat-affected zone. The feature dataset includes: hot crack features, porosity features, and grain features.

[0016] Apply feature recognition models and label each feature recognition model as an independent feature model to obtain the first brittleness recognition value qx1, the second brittleness recognition value qx2 and the third brittleness recognition value qx3;

[0017] The first embrittlement identification value qx1, the second embrittlement identification value qx2, and the third embrittlement identification value qx3 are compared with the defect threshold. If the value exceeds the defect threshold, a corresponding embrittlement detection record is generated; otherwise, no processing is performed.

[0018] The specific steps of applying the feature recognition model include: S21, obtaining the feature detection records of each weld of all components to be detected before the current time in the system, obtaining the heat-affected zone features in the feature detection records, establishing a feature dataset, and establishing a recognition model for each feature through a convolutional neural network (CNN);

[0019] S22. After training and validating the feature recognition model using the feature dataset, the hot crack features, porosity features, and grain features are input into the feature recognition model for analysis.

[0020] S23. Using an edge detection algorithm, identify the edge features of hot cracks in the heat-affected zone (HAZ) of the microscopic image, separate the hot crack region, and further optimize the extraction of the hot crack region through morphological analysis. Then, use image analysis software to count the separated hot crack regions, statistically determine the number of hot cracks in the image, and calculate the first embrittlement identification value qx1 of the HAZ using the following summation formula: In the formula, N is the number of detection records in the heat-affected zone, and C i The number of thermal cracks extracted in the i-th detection;

[0021] S24. The edge features of bubbles in the heat-affected zone of the microscopic image are identified using an edge detection algorithm to separate the bubble region. After further morphological optimization of the bubble region extraction, image analysis software is used to count the separated bubbles, and the number of bubbles in the image is statistically analyzed. The second embrittlement identification value qx2 of the heat-affected zone is calculated using the following summation formula: In the formula, N is the number of detection records in the heat-affected zone, and Q i The number of bubbles extracted in the i-th detection;

[0022] S25. Grain features of the heat-affected zone in the microscopic image are identified using an edge detection algorithm. Threshold segmentation is applied to extract the grains from the background. Each extracted grain is processed to obtain the total number of grains M. Contour detection is performed on the boundary of each grain, and the boundary lengths of all grains are calculated and summed to obtain the total grain length. Based on the total grain length and the total number of grains, the third embrittlement identification value qx3 is calculated using the following formula: Among them, L i Let be the boundary length of the i-th grain.

[0023] Preferably, the first embrittlement identification value qx1, the second embrittlement identification value qx2, and the third embrittlement identification value qx3 are compared with the defect threshold respectively. When the first embrittlement identification value qx1 is greater than the defect threshold, a hot crack embrittlement detection record is generated, indicating that there is hot crack embrittlement in the heat-affected zone during the welding process, and locating the hot crack embrittlement location.

[0024] When the second embrittlement identification value qx2 is greater than the defect threshold, a porosity embrittlement detection record is generated; this indicates that porosity embrittlement exists in the heat-affected zone during welding and the location of the embrittlement is identified.

[0025] When the third embrittlement identification value qx3 is greater than the defect threshold, a crystal coarsening detection record is generated; this indicates that there is crystal coarsening in the heat-affected zone during the welding process, and the location of the crystal coarsening is located.

[0026] The first embrittlement identification value qx1, the second embrittlement identification value qx2, and the third embrittlement identification value qx3 are sorted by size to form a sequence for extracting defects in the location of several heat-affected zones.

[0027] Preferably, the first embrittlement identification value qx1, the second embrittlement identification value qx2, and the third embrittlement identification value qx3 of each heat-affected zone are correlated, and the embrittlement risk coefficient Rch is obtained using the following formula: α, b, and d represent the weights of the first embrittlement identification value qx1, the second embrittlement identification value qx2, and the third embrittlement identification value qx3, respectively, with α = 0.3, b = 0.45, and d = 0.25. The embrittlement risk coefficient Rch is compared with the embrittlement threshold to determine the relative impact of each embrittlement phenomenon and obtain a comprehensive analysis result. If the embrittlement risk coefficient Rch is higher than the embrittlement threshold, the current heat-affected zone is marked as an embrittlement risk area, and the heat-affected zone with an embrittlement risk coefficient Rch lower than or equal to the embrittlement threshold is marked as a non-embrittlement stable area.

[0028] Preferably, the detection records of the same heat-affected zone characteristics of each component under test are marked as the first identical detection record. The detection results in the first identical detection record are counted, and records with the same result are marked as identical result records. The number and area of ​​the non-embrittled stable region of adjacent detection records in each group of identical result records are multiplied to obtain the total area Nns of the non-embrittled stable region. The stable area ratio Pwd is calculated according to the ratio of the total area Nns of the non-embrittled stable region to the total area A of the current component under test.

[0029] Preferably, the detection records with the same heat-affected zone characteristics for each component to be tested are marked as the second identical detection record. The detection results in the second identical detection record are counted, and records with the same result are marked as identical result records. For the embrittlement risk areas of adjacent detection records in each group of identical result records, the product calculation is performed to obtain the total area Nnx of the embrittlement risk area. Then, the embrittlement area ratio value Pfx is calculated according to the ratio of the total area Nnx of the embrittlement risk area to the total area A of the current component to be tested.

[0030] Preferably, when the stable area ratio Pwd is greater than the stability threshold, it indicates that the quality of the assembled steel structure component after the welding process is qualified, and a first qualified mark is generated; when the stable area ratio Pwd is less than or equal to the stability threshold, it indicates that the assembled steel structure is insufficient to support the entire assembled steel structure component at the connection point, and a first reinforcement strategy is generated, including: setting additional triangular reinforcement structures on the two adjacent sides of the welded joint, wherein the area of ​​the triangular reinforcement structures covers at least 3% of the total area A of the component to be tested;

[0031] When the embrittlement area ratio Pfx is greater than the embrittlement threshold, it indicates that the quality of the steel structure component of the assembly is unqualified after the welding process. A second repair strategy is generated, which includes: locating the hot crack embrittlement detection records, porosity embrittlement detection records, and crystal coarsening detection records of the steel structure component of the assembly, and filling the hot cracks and porosity embrittlement locations with 5% adhesive; filling the crystal coarsening locations with 5% alloy powder, followed by grinding and cutting, and then applying a protective coating; when the embrittlement area ratio Pfx is less than or equal to the embrittlement threshold, a second qualified mark is generated.

[0032] Preferably, after the first reinforcement strategy and the second repair strategy are implemented, steps S1-S4 are repeated. If the first and second qualification marks are still not generated simultaneously, in the first reinforcement strategy, the area of ​​the triangular reinforcement structure covering the total area A of the component to be tested is increased by at least 5%, and in the second repair strategy, the adhesive used to fill the hot cracks is increased from 5% to 10%, and the amount of alloy powder used to fill the crystal coarsening positions is increased from 5% to 10%. Only after all the steel structure components of the assembly have obtained the first and second qualification marks can they be shipped out as qualified products.

[0033] Preferably, if the first reinforcement strategy and the second repair strategy are repeated after the first reinforcement strategy and the second repair strategy are implemented, and if any heat-affected zone of the steel structure component of the assembly is repaired more than 3 times, it indicates that there is a risk of material stress fatigue, and the steel structure component of the assembly is disassembled and recycled.

[0034] This invention provides a method for quality inspection of prefabricated steel structure components based on image data. It has the following beneficial effects:

[0035] (1) By conducting detailed image analysis of the heat-affected zone, microscopic defects, such as embrittlement, can be effectively identified. Compared with traditional detection techniques, this method has higher sensitivity and accuracy, thus ensuring the quality of welded joints. The identification and assessment of embrittlement helps to discover potential structural risks in advance and reduce the probability of structural failure. By accurately marking the embrittled area, targeted measures can be taken during subsequent processing and use to improve the overall structural safety. By establishing unique identification numbers and size classifications, the components to be inspected can be quickly located, reducing the time cost of inspection and processing. This efficient inspection process helps to shorten the construction cycle and reduce the input of human resources. By comprehensively evaluating the stable area ratio value Pwd and the embrittled area ratio value Pfx, a multi-dimensional quality judgment standard is constructed. This dual verification mechanism can more comprehensively reflect the actual quality status of the components and ensure the effective screening of qualified products.

[0036] (2) Through systematic embrittlement characteristic detection of the heat-affected zone, the first embrittlement identification value qx1, the second embrittlement identification value qx2, and the third embrittlement identification value qx3 reflect the severity of hot cracking, porosity, and grain coarsening, respectively. This precise quantitative assessment enables engineers to identify potential material weakening and performance degradation during welding, providing a scientific basis for subsequent repair or reinforcement decisions. Multi-dimensional analysis of the embrittlement identification values ​​allows for the timely detection and marking of potential embrittlement risk areas. This process significantly enhances product reliability and safety, ensuring that prefabricated steel structure components can withstand expected loads in practical applications and reducing the probability of failure. By identifying and recording embrittlement phenomena in advance, targeted measures can be taken during subsequent processing or re-welding to reduce the risk of further damage to the heat-affected zone. For example, if severe embrittlement is found in a certain area, different welding processes or reinforcement strategies can be selected to avoid larger quality problems caused by porosity or hot cracking.

[0037] A comprehensive assessment of multiple heat-affected zones (HAZs) allows for the assessment of the overall health of the welded structure. This holistic monitoring helps identify potential embrittlement issues, extending beyond individual weld points to encompass the entire component, thereby improving overall structural safety. By comprehensively analyzing the embrittlement risk of different HAZs, areas poised for problems can be identified more sensitively. This sensitivity enables early detection of microscopic defects, preventing larger-scale structural damage due to delayed intervention.

[0038] (3) By calculating the stable area ratio (Pwd) and the embrittled area ratio (Pfx), the consistency of these data can be mutually verified. If the calculated results of the two ratios are similar, it indicates the accuracy and reliability of the data, which provides a solid data foundation for subsequent analysis. Through mutual verification, management can gain a more comprehensive understanding of the health status of the component to be inspected. For example, if the stable area ratio is high and the embrittled area ratio is low, it indicates that the component is generally healthy, and vice versa, which can better guide maintenance and repair decisions. By considering the ratios of stability and embrittlement simultaneously, the safety risks of the component can be assessed more comprehensively.

[0039] (4) When the proportion of stable area is greater than the stability threshold, the component is marked as qualified, ensuring its load-bearing capacity after welding and reducing potential safety hazards. When the proportion of embrittled area exceeds the embrittlement threshold, a repair strategy is generated in a timely manner to prevent unqualified components from being put into use and to ensure the safety of the overall structure. Specific reinforcement or repair measures are generated based on different test results, such as setting additional triangular reinforcement structures at welded joints and filling hot cracks and pores, ensuring the pertinence and effectiveness of the treatment measures. During the reinforcement process, setting additional triangular reinforcement structures can effectively utilize materials and only needs to cover at least 3% of the total area of ​​the component to be tested, thus improving the load-bearing capacity of the structure without wasting materials. Through the implementation of test results and repair strategies, a dynamic feedback mechanism is formed, enabling subsequent welding quality control and test strategies to be continuously adjusted and optimized. Attached Figure Description

[0040] Figure 1 This is a schematic diagram illustrating the steps of the image data-based quality inspection method for prefabricated steel structure components according to the present invention. Detailed Implementation

[0041] 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.

[0042] Example 1

[0043] Please see Figure 1 This invention provides a method for quality inspection of prefabricated steel structure components based on image data, comprising the following steps:

[0044] S1. Size Classification: Each prefabricated steel structure component to be inspected after welding is assigned a unique identification number, and after identifying dimensional characteristics, it is classified by size to obtain several size groups; the size groups include:

[0045] The surface area of ​​the prefabricated steel structure component to be inspected after welding is less than 1m². 3 The components to be inspected are classified into the first small component group; the surface area of ​​the prefabricated steel structure components to be inspected after welding is equal to 1m². 3 And less than 10m 3 The components to be inspected are classified into the second medium-sized component group; the prefabricated steel structure components to be inspected after welding have a surface area greater than 10m². 3 The components to be tested are classified into the third large component group.

[0046] In practice, components of different sizes may require different equipment and human resources. Grouping helps managers better allocate resources, ensuring that the inspection of each component receives appropriate attention and handling. Size classification helps establish unified inspection standards, allowing different inspectors to evaluate according to the same criteria, thereby improving the consistency and reliability of inspection results. Inspection results for components in different size groups can be fed back into production and welding processes, allowing for timely adjustments to process parameters and optimization of the production flow to reduce subsequent quality problems.

[0047] S2. Image Recognition: Using automated welding inspection equipment and a computer-controlled scanning electron microscope, microscopic image data of each weld in the assembled steel structure components to be inspected after welding is collected in several size groups. These devices can ensure consistent inspection of each weld point under the same conditions. Each weld represents a heat-affected zone (HAZ). The HAZ features are extracted from the microscopic image data. Each time the HAZ of the steel structure component is identified, the HAZ is labeled and the corresponding microscopic image is obtained. The microscopic images are then used for feature recognition according to the order of HAZ feature extraction. After the feature recognition of each HAZ is completed, a feature dataset is established, and the following three embrittlement identification values ​​are obtained: first embrittlement identification value qx1, second embrittlement identification value qx2, and third embrittlement identification value qx3.

[0048] Based on microscopic images, a feature dataset is obtained for several heat-affected zones, including hot crack features, porosity features, and grain features. By acquiring these features, the quality of the welded joint can be comprehensively evaluated. Each feature reflects a different type of defect. Through comprehensive analysis of these features, the health status of the welded area can be more accurately determined, thereby effectively reducing structural risks caused by quality problems.

[0049] Porosity is also an internal defect in materials, and its presence can lead to stress concentration. When a material is subjected to external loads such as tension or compression, the stress near the pores is higher than in areas farther away, increasing the risk of fracture. The presence of hot cracks and porosity can reduce the fatigue life of materials under repeated loading. Under fatigue loading, hot cracks and porosity can trigger crack propagation, leading to embrittlement of the material under fatigue stress.

[0050] Hot cracks are defects in materials that can create stress concentrations under stress. The stress at the tip of a hot crack is much higher than in the surrounding area, making the material more prone to fracture under stress. This stress concentration makes the material more susceptible to brittle fracture when subjected to external loads.

[0051] According to the Hall-Petch criterion, smaller grain size is generally associated with higher material strength and toughness. Grain enlargement due to grain coarsening reduces the yield strength and tensile strength of the material, making it more susceptible to brittle fracture. In coarsened grains, the reduced number of grain boundaries can lead to stress concentration. Under stress, these stress concentration areas are more prone to cracking, thus inducing embrittlement. During welding, the material in the weld heat-affected zone undergoes rapid heating and cooling, resulting in grain coarsening. This region is typically associated with a higher risk of embrittlement, especially at the weld joint and in adjacent areas.

[0052] Apply feature recognition models and label each feature recognition model as an independent feature model to obtain the first brittleness recognition value qx1, the second brittleness recognition value qx2 and the third brittleness recognition value qx3;

[0053] The first embrittlement identification value qx1, the second embrittlement identification value qx2, and the third embrittlement identification value qx3 are compared with the defect threshold. If the value exceeds the defect threshold, a corresponding embrittlement detection record is generated; otherwise, no processing is performed.

[0054] The specific steps of applying the feature recognition model include: S21, acquiring feature detection records for each weld of all components to be inspected up to the current time in the system, obtaining the heat-affected zone features from the feature detection records, establishing a feature dataset, and building a recognition model for each feature using a convolutional neural network (CNN). Traditional detection methods often perform well in identifying macroscopic defects, but are less effective at identifying microscopic defects (such as embrittlement of the heat-affected zone). Applying a convolutional neural network (CNN) to build a feature recognition model can improve the detection sensitivity of subtle defects, promptly identify potential embrittlement risks, and provide important basis for subsequent processing.

[0055] S22. After training and validating the feature recognition model using the feature dataset, hot crack features, porosity features, and grain features are input into the feature recognition model for analysis. The system can analyze the feature detection records of all components to be inspected up to the current time, achieving real-time monitoring. This dynamic feedback mechanism enables rapid identification and handling of defects during the welding process, reducing subsequent repair costs and improving production efficiency.

[0056] S23. Using an edge detection algorithm, identify the edge features of hot cracks in the heat-affected zone (HAZ) of the microscopic image, separate the hot crack region, and further optimize the extraction of the hot crack region through morphological analysis. Then, use image analysis software to count the separated hot crack regions, statistically determine the number of hot cracks in the image, and calculate the first embrittlement identification value qx1 of the HAZ using the following summation formula: In the formula, N is the number of detection records in the heat-affected zone, and C i The number of thermal cracks extracted in the i-th detection;

[0057] S24. The edge features of bubbles in the heat-affected zone of the microscopic image are identified using an edge detection algorithm to separate the bubble region. After further morphological optimization of the bubble region extraction, image analysis software is used to count the separated bubbles, and the number of bubbles in the image is statistically analyzed. The second embrittlement identification value qx2 of the heat-affected zone is calculated using the following summation formula: In the formula, N is the number of detection records in the heat-affected zone, and Q i The number of bubbles extracted in the i-th detection;

[0058] S25. Grain features of the heat-affected zone in the microscopic image are identified using an edge detection algorithm. Threshold segmentation is applied to extract the grains from the background. Each extracted grain is processed to obtain the total number of grains M. Contour detection is performed on the boundary of each grain, and the boundary lengths of all grains are calculated and summed to obtain the total grain length. Based on the total grain length and the total number of grains, the third embrittlement identification value qx3 is calculated using the following formula: Among them, L i Let be the boundary length of the i-th grain.

[0059] S3. Embrittlement determination: Based on the first embrittlement identification value qx1, the second embrittlement identification value qx2 and the third embrittlement identification value qx3, an embrittlement risk coefficient Rch is constructed. Based on the comparison between the embrittlement risk coefficient Rch and the embrittlement threshold, the relative influence of each embrittlement phenomenon is determined, and the embrittled area and the non-embrittled area are marked.

[0060] S4. Dual verification judgment: Statistically calculate the stable area ratio value Pwd and the embrittled area ratio value Pfx, and evaluate whether each component to be tested is a qualified product based on the stable area ratio value Pwd and the embrittled area ratio value Pfx.

[0061] In this embodiment, detailed image analysis of the heat-affected zone effectively identifies microscopic defects, such as embrittlement. Compared to traditional detection techniques, this method offers higher sensitivity and accuracy, ensuring the quality of welded joints. The identification and assessment of embrittlement helps to detect potential structural risks early, reducing the probability of structural failure. Accurate marking of embrittled areas allows for targeted measures to be taken during subsequent processing and use, improving the overall structural safety. Establishing unique identification numbers and size classifications enables rapid location of components to be inspected, reducing inspection and processing time costs. This efficient inspection process helps shorten construction cycles and reduce human resource investment. A multi-dimensional quality judgment standard is constructed through a comprehensive evaluation of the stable area ratio (Pwd) and the embrittled area ratio (Pfx). This dual verification mechanism more comprehensively reflects the actual quality status of components, ensuring the effective screening of qualified products.

[0062] Example 2

[0063] This embodiment is an explanation of Embodiment 1. Specifically, the first embrittlement identification value qx1, the second embrittlement identification value qx2, and the third embrittlement identification value qx3 are compared with the defect threshold. When the first embrittlement identification value qx1 is greater than the defect threshold, a hot crack embrittlement detection record is generated, indicating that there is a hot crack embrittlement phenomenon in the heat-affected zone during the welding process, and locating the hot crack embrittlement location.

[0064] When the second embrittlement identification value qx2 is greater than the defect threshold, a porosity embrittlement detection record is generated; this indicates that porosity embrittlement exists in the heat-affected zone during welding and the location of the embrittlement is identified.

[0065] When the third embrittlement identification value qx3 is greater than the defect threshold, a crystal coarsening detection record is generated; this indicates that there is crystal coarsening in the heat-affected zone during the welding process, and the location of the crystal coarsening is located.

[0066] The first embrittlement identification value qx1, the second embrittlement identification value qx2, and the third embrittlement identification value qx3 are sorted by size to form a sequence for extracting defects in the location of several heat-affected zones.

[0067] In this embodiment, by systematically detecting the embrittlement characteristics of the welded joint and its heat-affected zone, the obtained first embrittlement identification value qx1, second embrittlement identification value qx2, and third embrittlement identification value qx3 reflect the severity of hot cracking, porosity, and grain coarsening, respectively. This precise quantitative assessment allows engineers to identify potential material weakening and performance degradation that may occur during the welding process, providing a scientific basis for subsequent repair or reinforcement decisions. Multi-dimensional analysis of the embrittlement identification values ​​enables the timely detection and marking of potential embrittlement risk areas. This process significantly enhances product reliability and safety, ensuring that prefabricated steel structure components can withstand expected loads in practical applications and reducing the probability of failure. By identifying and recording embrittlement phenomena in advance, targeted measures can be taken during subsequent processing or re-welding to reduce the risk of further damage to the heat-affected zone. For example, if severe embrittlement is found in a certain area, different welding processes or reinforcement strategies can be selected to avoid larger quality problems caused by porosity or hot cracking.

[0068] Example 3

[0069] This embodiment is an explanation of Embodiment 2. Specifically, the first embrittlement identification value qx1, the second embrittlement identification value qx2, and the third embrittlement identification value qx3 of each heat-affected zone are correlated, and the embrittlement risk coefficient Rch is obtained using the following formula: α, b, and d represent the weights of the first embrittlement identification value qx1, the second embrittlement identification value qx2, and the third embrittlement identification value qx3, respectively, with α = 0.3, b = 0.45, and d = 0.25. The embrittlement risk coefficient Rch is compared with the embrittlement threshold to determine the relative impact of each embrittlement phenomenon and obtain a comprehensive analysis result. If the embrittlement risk coefficient Rch is higher than the embrittlement threshold, the current heat-affected zone is marked as an embrittlement risk area, and the heat-affected zone with an embrittlement risk coefficient Rch lower than or equal to the embrittlement threshold is marked as a non-embrittlement stable area.

[0070] In this embodiment, a comprehensive assessment of multiple heat-affected zones (HAZs) allows for the acquisition of the overall health status of the welded structure. This comprehensive monitoring helps identify potential embrittlement problems, extending beyond individual weld points to encompass the entire component, thereby improving overall structural safety. By comprehensively analyzing the embrittlement risk of different HAZs, areas poised for problems can be identified more sensitively. This sensitivity enables early detection of microscopic defects, preventing larger-scale structural damage due to delayed intervention.

[0071] Example 4

[0072] This embodiment is an explanation of Embodiment 3. Specifically, the detection records of the same heat-affected zone characteristics of each component to be tested are marked as the first identical characteristic detection record. The detection results in the first identical characteristic detection record are counted, and records with the same result are marked as identical result records. The number and area of ​​the non-embrittled stable region of adjacent detection records in each group of identical result records are multiplied to obtain the total area Nns of the non-embrittled stable region. The stable area ratio value Pwd is calculated according to the ratio of the total area Nns of the non-embrittled stable region to the total area A of the current component to be tested.

[0073] The detection records with the same heat-affected zone characteristics for each component under test are marked as the second identical detection record. The detection results in the second identical detection record are counted, and records with the same result are marked as identical result records. For the embrittlement risk areas of adjacent detection records in each group of identical result records, the product is calculated to obtain the total area Nnx of the embrittlement risk area. Then, the embrittlement area ratio Pfx is calculated based on the total area Nnx of the embrittlement risk area and the total area A of the current component under test.

[0074] In this embodiment, the consistency of these data can be verified by calculating the stable area ratio (Pwd) and the embrittlement area ratio (Pfx). If the calculated results of the two ratios are similar, it indicates the accuracy and reliability of the data, providing a solid data foundation for subsequent analysis. Through mutual verification, management can gain a more comprehensive understanding of the health status of the component under inspection. For example, a high stable area ratio and a low embrittlement area ratio indicate that the component is generally healthy, and vice versa, better guiding maintenance and repair decisions. By considering both the stable and embrittlement ratios simultaneously, the safety risks of the component can be assessed more comprehensively. For example, a high embrittlement area ratio without a corresponding decrease in the stable area ratio may indicate potential structural problems, alerting relevant personnel to take action. If there is a significant difference between the stable and embrittlement area ratios, more dynamic and targeted maintenance strategies can be developed based on the analysis results. For example, for components with a high embrittlement ratio, the frequency of maintenance and inspection can be increased to ensure safety. By comprehensively analyzing these two ratios, the accuracy and timeliness of fault prediction can be improved. If the proportion of embrittled area increases while the proportion of stable area decreases, it can serve as a warning signal of a fault, prompting engineers to conduct timely inspections and take appropriate action.

[0075] Example 5

[0076] This embodiment is an explanation of Embodiment 4. Specifically, when the stable area ratio value Pwd is greater than the stability threshold, it indicates that the quality of the assembled steel structure component after welding is qualified, and a first qualified mark is generated. When the stable area ratio value Pwd is less than or equal to the stability threshold, it indicates that the assembled steel structure is insufficient to bear the weight of the entire assembled steel structure component at the connection point, and a first reinforcement strategy is generated, including: setting additional triangular reinforcement structures on the two adjacent sides of the welded joint, the area of ​​which covers at least 3% of the total area A of the component under test; using a triangular shape for reinforcement is because the triangular structure has good stability and load-bearing capacity in mechanics, and can effectively disperse the force applied to the welded joint. The area of ​​the reinforcement structure should cover at least 3% of the total area A of the component under test. This means that the area of ​​the reinforcement structure must reach a certain proportion relative to the area of ​​the entire component to ensure that its reinforcement effect is obvious and effective. For example, if the total area of ​​the component under test is 100m²... 2 Therefore, the area of ​​the reinforced structure needs to be at least 3m². 2 .

[0077] When the embrittlement area ratio Pfx is greater than the embrittlement threshold, it indicates that the quality of the steel structure component of the assembly is unqualified after the welding process. A second repair strategy is generated, which includes: locating the hot crack embrittlement detection records, porosity embrittlement detection records, and crystal coarsening detection records of the steel structure component of the assembly, and filling the hot cracks and porosity embrittlement locations with 5% adhesive; filling the crystal coarsening locations with 5% alloy powder, followed by grinding and cutting, and then applying a protective coating; when the embrittlement area ratio Pfx is less than or equal to the embrittlement threshold, a second qualified mark is generated.

[0078] In this embodiment, when the stable area ratio is greater than the stability threshold, the component is marked as qualified, ensuring its load-bearing capacity after welding and reducing potential safety hazards. Conversely, when the embrittlement area ratio exceeds the embrittlement threshold, a repair strategy is generated promptly to prevent unqualified components from being used, ensuring the overall structural safety. Specific reinforcement or repair measures are generated based on different test results, such as setting additional triangular reinforcement structures at weld joints and filling hot cracks and porosity, ensuring the targeted and effective nature of the treatment measures. During reinforcement, setting additional triangular reinforcement structures effectively utilizes material and only needs to cover at least 3% of the total area of ​​the component under test, thus improving the structural load-bearing capacity without wasting material. Through the implementation of test results and repair strategies, a dynamic feedback mechanism is formed, enabling continuous adjustment and optimization of subsequent welding quality control and testing strategies.

[0079] Example 6

[0080] This embodiment is an explanation of embodiment 5. Specifically, after the first reinforcement strategy and the second repair strategy are implemented, steps S1-S4 are repeated. If the first and second qualification marks are still not generated simultaneously, in the first reinforcement strategy, the area of ​​the triangular reinforcement structure is increased to cover at least 5% of the total area A of the component to be tested, and in the second repair strategy, the adhesive used to fill the hot cracks is increased from 5% to 10%, and the amount of alloy powder used to fill the crystal coarsening positions is increased from 5% to 10%. Only after all the steel structure components of the assembly have obtained the first and second qualification marks can they be shipped out as qualified products.

[0081] If, after implementing the first reinforcement strategy and the second repair strategy, the first reinforcement strategy and the second repair strategy in steps S1-S4 are repeated, and if any heat-affected zone of the steel structure component of the assembly is repaired more than 3 times, it indicates that there is a risk of material stress fatigue, and the steel structure component of the assembly is disassembled and recycled.

[0082] In this embodiment, by setting additional triangular reinforcement structures on two adjacent sides of the welded joint, the strength of the welded joint can be effectively enhanced, and stress concentration can be reduced. As the area of ​​the triangular reinforcement increases to at least 5% of the total area, the stability of the structure is further improved, ensuring that it can withstand higher loads and stresses during use, thereby greatly reducing the risk of welded joint failure. In the second repair strategy, for filling hot cracks and pores, the proportion of filling material used is gradually increased from 5% to 10%, which helps to improve the filling effect and ensure that hot cracks and pores are fully repaired. By using a higher proportion of adhesive and alloy powder, defects can be filled more effectively, and the mechanical properties of the repaired area can be improved, making it closer to the strength and toughness of the original material. By continuously repeating steps S1-S4, the system can re-evaluate the repaired heat-affected zone. This feedback mechanism ensures the effectiveness of each repair, timely detection and correction of potential problems. If the first and second qualification marks are not generated simultaneously, it indicates that the repair effect has not met expectations, and the strategy is further adjusted and incremental repairs are carried out to ensure the qualification of product quality. During the repair process, if a heat-affected zone is found to have been repaired more than three times, it indicates a potential risk of material stress fatigue in that area. This mechanism enables the timely identification and removal of potentially hazardous products, preventing safety hazards caused by quality defects. This preventative maintenance measure significantly reduces the risk of accidents, ensuring safety and reliability during use. Only after all heat-affected zones have received the first and second acceptance marks will the assembled steel structural component be considered a qualified product and released from the warehouse. This rigorous quality control process ensures that products undergo thorough testing and repair before delivery, promoting improved safety and reliability in use.

[0083] The threshold is set to facilitate comparison. The size of the threshold depends on the amount of sample data and the number of bases set by those skilled in the art for each set of sample data; as long as it does not affect the ratio between the parameter and the quantized value, it is acceptable.

[0084] The above formulas are all derived from software simulation using a large amount of data and are selected to be close to the actual values. The coefficients in the formulas are set by those skilled in the art according to the actual situation. The above description is only a preferred embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any equivalent substitutions or changes made by those skilled in the art within the technical scope disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the protection scope of the present invention.

Claims

1. A method for quality inspection of prefabricated steel structure components based on image data, characterized in that, Includes the following steps: S1. Size Classification: Assign a unique identification number to each prefabricated steel structure component to be inspected after welding is completed, and classify it by size after identifying size characteristics to obtain several size groups; S2. Image Recognition: Collect microscopic image data of each weld joint of the prefabricated steel structure components to be inspected after welding in several size groups. Each weld joint represents a heat-affected zone. Extract the heat-affected zone features from the microscopic image data. When identifying the heat-affected zone of the steel structure component each time, the heat-affected zone is labeled and the corresponding microscopic image is obtained. Perform feature recognition on the microscopic images according to the feature extraction order of the heat-affected zone. After completing the recognition of the features of each heat-affected zone, establish a feature dataset and obtain: the first embrittlement recognition value qx1, the second embrittlement recognition value qx2, and the third embrittlement recognition value qx3. S3. Embrittlement determination: Based on the first embrittlement identification value qx1, the second embrittlement identification value qx2 and the third embrittlement identification value qx3, an embrittlement risk coefficient Rch is constructed. Based on the comparison between the embrittlement risk coefficient Rch and the embrittlement threshold, the relative influence of each embrittlement phenomenon is determined, and the embrittled area and the non-embrittled area are marked. S4. Dual verification judgment: Statistically calculate the stable area ratio value Pwd and the embrittled area ratio value Pfx, and evaluate whether each component to be tested is a qualified product based on the stable area ratio value Pwd and the embrittled area ratio value Pfx.

2. The method for quality inspection of prefabricated steel structure components based on image data according to claim 1, characterized in that, Grouped by several dimensions, including: The surface area of ​​the prefabricated steel structure component to be inspected after welding is less than 1m². 3 The components to be tested are classified into the first small component group; The surface area of ​​the prefabricated steel structure component to be inspected after welding is equal to 1m². 3 and less than 10m 3 The components to be tested are classified into the second medium-sized component group; The surface area of ​​the prefabricated steel structure components to be inspected after welding is greater than 10m². 3 The components to be tested are classified into the third large component group.

3. The method for quality inspection of prefabricated steel structure components based on image data according to claim 1, characterized in that, Feature recognition of heat-affected zone images specifically includes: obtaining a feature dataset of several heat-affected zone features based on microscopic images, including: hot crack features, porosity features, and grain features; Apply feature recognition models and label each feature recognition model as an independent feature model to obtain the first brittleness recognition value qx1, the second brittleness recognition value qx2 and the third brittleness recognition value qx3; The first embrittlement identification value qx1, the second embrittlement identification value qx2, and the third embrittlement identification value qx3 are compared with the defect threshold. If the value exceeds the defect threshold, a corresponding embrittlement detection record is generated; otherwise, no processing is performed. The specific steps of applying the feature recognition model include: S21, obtaining the feature detection records of each weld of all components to be detected before the current time in the system, obtaining the heat-affected zone features in the feature detection records, establishing a feature dataset, and establishing a recognition model for each feature through a convolutional neural network (CNN); S22. After training and validating the feature recognition model using the feature dataset, the hot crack features, porosity features, and grain features are input into the feature recognition model for analysis. S23. Using an edge detection algorithm, identify the edge features of hot cracks in the heat-affected zone (HAZ) of the microscopic image, separate the hot crack region, and further optimize the extraction of the hot crack region through morphological analysis. Then, use image analysis software to count the separated hot crack regions, statistically determine the number of hot cracks in the image, and calculate the first embrittlement identification value qx1 of the HAZ using the following summation formula: In the formula, N is the number of detection records for the heat-affected zone, and C i The number of thermal cracks extracted in the i-th detection; S24. The bubble features in the heat-affected zone of the microscopic image are identified using an edge detection algorithm, and the bubble region is separated. After further morphological optimization of the bubble region extraction, image analysis software is used to count the separated bubbles, and the number of bubbles in the image is statistically analyzed. The second embrittlement identification value qx2 of the heat-affected zone is calculated using the following summation formula: In the formula, N is the number of detection records in the heat-affected zone, and Q i The number of bubbles extracted in the i-th detection; S25. Grain features of the heat-affected zone in the microscopic image are identified using an edge detection algorithm. Threshold segmentation is applied to extract the grains from the background. Each extracted grain is processed to obtain the total number of grains M. Contour detection is performed on the boundary of each grain, and the boundary lengths of all grains are calculated and summed to obtain the total grain length. Based on the total grain length and the total number of grains, the third embrittlement identification value qx3 is calculated using the following formula: Among them, L i Let be the boundary length of the i-th grain.

4. The method for quality inspection of prefabricated steel structure components based on image data according to claim 3, characterized in that, The first embrittlement identification value qx1, the second embrittlement identification value qx2, and the third embrittlement identification value qx3 are compared with the defect threshold. If the first embrittlement identification value qx1 is greater than the defect threshold, a hot crack embrittlement detection record is generated, indicating that hot crack embrittlement exists in the heat-affected zone during welding, and the location of the hot crack embrittlement is located. If the second embrittlement identification value qx2 is greater than the defect threshold, a porosity embrittlement detection record is generated, indicating that porosity embrittlement exists in the heat-affected zone during welding, and the location of the porosity embrittlement is located. If the third embrittlement identification value qx3 is greater than the defect threshold, a crystal coarsening detection record is generated, indicating that crystal coarsening exists in the heat-affected zone during welding, and the location of the crystal coarsening is located. The first embrittlement identification value qx1, the second embrittlement identification value qx2, and the third embrittlement identification value qx3 are sorted by size to form a sequence for extracting defects in the location of several heat-affected zones.

5. The method for quality inspection of prefabricated steel structure components based on image data according to claim 4, characterized in that, The first embrittlement identification value qx1, the second embrittlement identification value qx2, and the third embrittlement identification value qx3 of each heat-affected zone are correlated, and the embrittlement risk coefficient Rch is obtained using the following formula: α, b, and d represent the weight values ​​of the first embrittlement identification value qx1, the second embrittlement identification value qx2, and the third embrittlement identification value qx3, respectively. The embrittlement risk coefficient Rch is compared with the embrittlement threshold to determine the relative influence of each embrittlement phenomenon and obtain a comprehensive analysis result. If the embrittlement risk coefficient Rch is higher than the embrittlement threshold, the current heat-affected zone is marked as an embrittlement risk area, and the heat-affected zone with an embrittlement risk coefficient Rch lower than or equal to the embrittlement threshold is marked as a non-embrittlement stable area.

6. The method for quality inspection of prefabricated steel structure components based on image data according to claim 1, characterized in that, The detection records with the same heat-affected zone characteristics for each component under test are marked as the first identical detection record. The detection results in the first identical detection record are statistically analyzed, and records with the same result are marked as identical result records. The number and area of ​​non-embrittled stable regions in adjacent detection records within each group of identical result records are multiplied to obtain the total area Nns of non-embrittled stable regions. The stable area ratio Pwd is then calculated based on the ratio of the total area Nns of non-embrittled stable regions to the total area A of the current component under test.

7. The method for quality inspection of prefabricated steel structure components based on image data according to claim 6, characterized in that, The detection records with the same heat-affected zone characteristics for each component under test are marked as the second identical detection record. The detection results in the second identical detection record are counted, and records with the same result are marked as identical result records. For the embrittlement risk areas of adjacent detection records in each group of identical result records, the product is calculated to obtain the total area Nnx of the embrittlement risk area. Then, the embrittlement area ratio Pfx is calculated based on the total area Nnx of the embrittlement risk area and the total area A of the current component under test.

8. The method for quality inspection of prefabricated steel structure components based on image data according to claim 7, characterized in that, When the stable area ratio Pwd is greater than the stable threshold, it indicates that the quality of the steel structure component of the assembly is qualified after the welding process, and the first qualified mark is generated. When the stable area ratio Pwd is less than or equal to the stable threshold, it indicates that the steel structure of the assembly is insufficient to bear the weight of the entire steel structure component at the connection point. A first reinforcement strategy is generated, including: setting additional triangular reinforcement structures on the two adjacent sides of the welded joint, wherein the area of ​​the triangular reinforcement structures covers at least 3% of the total area A of the component to be tested. When the embrittlement area ratio Pfx is greater than the embrittlement threshold, it indicates that the quality of the steel structure component of the assembly is unqualified after the welding process. A second repair strategy is generated, which includes: locating the hot crack embrittlement detection records, porosity embrittlement detection records, and crystal coarsening detection records of the steel structure component of the assembly, and filling the hot cracks and porosity embrittlement locations with 5% adhesive; filling the crystal coarsening locations with 5% alloy powder, followed by grinding and cutting, and then applying a protective coating; when the embrittlement area ratio Pfx is less than or equal to the embrittlement threshold, a second qualified mark is generated.

9. The method for quality inspection of prefabricated steel structure components based on image data according to claim 8, characterized in that, After the first reinforcement strategy and the second repair strategy are implemented, steps S1-S4 are repeated. If the first and second qualification marks are still not generated simultaneously, in the first reinforcement strategy, the area of ​​the triangular reinforcement structure is increased to cover at least 5% of the total area A of the component to be tested, and in the second repair strategy, the adhesive used to fill the hot cracks is increased from 5% to 10%, and the amount of alloy powder used to fill the crystal coarsening areas is increased from 5% to 10%. Only after all the steel structure components of the assembly have obtained the first and second qualification marks can they be shipped out as qualified products.

10. The method for quality inspection of prefabricated steel structure components based on image data according to claim 9, characterized in that, If, after implementing the first reinforcement strategy and the second repair strategy, the first reinforcement strategy and the second repair strategy in steps S1-S4 are repeated, and if any heat-affected zone of the steel structure component of the assembly is repaired more than 3 times, it indicates that there is a risk of material stress fatigue, and the steel structure component of the assembly is disassembled and recycled.

Images