Bridge steel structure welding quality detection device and method

Abstract

The invention belongs to the technical field of steel structure welding detection, and discloses a bridge steel structure welding quality detection device and method, and the method comprises the steps that a route planning module plans an optimal flight path; the cruise monitoring module carries out all-directional welding quality detection on the bridge according to the planned route; the image processing and recognition module is used for carrying out preprocessing, feature extraction and recognition on the image and recognizing the welding quality problem of the bridge steel structure; the infrared data analysis and identification module is used for processing and analyzing infrared wave data collected by the infrared thermal imager and identifying the welding quality problem of the bridge steel structure; and the comprehensive evaluation module combines the results of the image processing identification module and the infrared data analysis identification module, and comprehensively evaluates the overall welding quality of the steel structure. According to the invention, efficient and comprehensive detection of the bridge steel structure can be realized, the collected data can be processed in real time through the image processing and infrared data analysis module, and the welding quality problem can be quickly and accurately identified.

Description

Technical Field

[0001] The present application relates to the technical field of steel structure welding detection, and more specifically, to a device and method for detecting the welding quality of a bridge steel structure. Background Art

[0002] The steel structure of the beam is a key component that supports and connects the various parts of the bridge. Its welding quality is directly related to the overall strength, stability and durability of the bridge. Therefore, the detection of steel structure welds is an indispensable and important part of bridge construction and maintenance. The purpose of weld detection is to timely discover and repair welding defects through detection, and prevent structural failure or safety accidents caused by weld quality problems.

[0003] The document with the prior art publication number CN118010854A provides a nondestructive testing device for the welding quality of a bridge steel structure and a method for using the same, which belongs to the technical field of steel structure welding detection. It includes a base; a fixed frame is installed on the top of the base; a slide groove is opened on the top of the fixed frame; a slider is clamped in the slider; a U-shaped seat is provided at the bottom of the slider; an ultrasonic probe is installed in the U-shaped seat; a walking mechanism is installed on the top of the slider; a fitting mechanism is installed between the U-shaped seat and the ultrasonic probe, and the fitting mechanism is connected to the walking mechanism; a coating mechanism is installed on the U-shaped seat, and the coating mechanism is connected to the fitting mechanism, and the walking mechanism, the fitting mechanism and the coating mechanism are linked to realize that the coupling agent is first applied to the weld, and then the ultrasonic probe is attached to the weld for detection, and after detecting a point, the probe is lifted up, moved a short distance along the weld and then moved down again for detection, so as to achieve the purpose of comprehensive detection of the weld, avoid wear of the ultrasonic probe during movement, and the coupling agent is used as a transmission medium during detection to improve the detection accuracy.

[0004] Although the above-mentioned prior art solutions can achieve relevant beneficial effects through the structure of the prior art, they still have the following defects: 1. The single welding defect detection method is not comprehensive and accurate enough for welding defect detection. 2. It is not suitable for quality inspection of bridge steel structure after on-site welding. Due to the height of the bridge steel structure and other reasons, the prior art cannot detect the quality of bridge on-site welding.

[0005] In view of this, we propose a device and method for detecting the welding quality of bridge steel structures. Summary of the invention

[0006] 1. Technical problems to be solved

[0007] The purpose of the present application is to provide a device and method for detecting the welding quality of bridge steel structures, which solves the technical problems raised in the above-mentioned background technology, realizes efficient and comprehensive detection of bridge steel structures, and avoids omissions and subjective judgment errors that may exist in traditional manual detection; through image processing and infrared data analysis modules, the collected data can be processed in real time, and welding quality problems can be quickly identified, so that problems can be discovered and reported in time; combined with the multi-dimensional detection means of high-definition cameras and infrared thermal imagers, the welding quality information of bridge steel structures can be captured more comprehensively, and various data can be integrated and analyzed through a comprehensive evaluation module, thereby improving the comprehensiveness and accuracy of the evaluation. Technical effect.

[0008] 2. Technical solution

[0009] The technical solution of the present application provides a bridge steel structure welding quality detection device, comprising:

[0010] Data acquisition module: collects the design drawings of the bridge steel structure, including the material, specifications and welding requirements (welding position, welding method and weld thickness, etc.) of the steel structure, and collects images of welding defects; annotates the data as a reference sample;

[0011] Route planning module: Based on the design drawings, GIS and CAD technology are used to build a three-dimensional model of the bridge steel structure to accurately simulate its spatial layout. Considering factors such as the flight speed, altitude, stability, and endurance of the drone, combined with the distribution of obstacles in the three-dimensional model (such as bridge support structures, cables, etc.), the algorithm calculates the optimal flight path to ensure that the drone can complete the inspection task efficiently and safely.

[0012] Cruise monitoring module: including drones, infrared thermal imagers, high-definition cameras and LED lights. According to the planned route, the drones carry high-definition cameras and infrared thermal imagers to conduct a full range of welding quality inspections on the bridges.

[0013] Image processing and recognition module: pre-processes, extracts features and recognizes images collected by high-definition cameras, and promptly identifies welding quality problems of bridge steel structures, including cracks, lack of fusion, pores, slag inclusions and deformation; identifies the size of steel materials, compares and analyzes them with the specified size, and promptly identifies abnormal situations where the material size does not meet the requirements;

[0014] Infrared data analysis and identification module: Process and analyze the infrared wave data collected by the infrared thermal imager to identify welding quality problems of bridge steel structures, including cracks, lack of fusion, pores, slag inclusions and deformation;

[0015] Comprehensive evaluation module: Combines the analysis results of the image processing recognition module and the infrared data analysis recognition module to comprehensively evaluate the overall welding quality of the steel structure. Based on factors such as defect type, location, and severity, the impact of welding quality on the overall safety of the bridge is evaluated.

[0016] Alarm module: including an alarm, which will sound an alarm in time when welding quality problems are detected;

[0017] Control center: network connected with data collection module, route planning module, cruise monitoring module, image processing and recognition module, ultrasonic data analysis and recognition module, comprehensive evaluation module and alarm module.

[0018] As an optional solution of the present invention, the route planning module plans the flight route of the drone, including the following steps:

[0019] 1. Build a 3D model: Take the design drawings of the bridge steel structure as input. The drawings contain detailed information such as the size, shape, location, and welding requirements of the steel structure. Based on the design drawings, use geographic information system (GIS) and computer-aided design (CAD) technology to build a 3D model of the bridge steel structure. Through 3D modeling, the spatial layout of the bridge steel structure can be accurately simulated, including key parts such as the bridge's support structure, cables, and welded joints.

[0020] 2. Set the flight parameters of the drone: Set the appropriate flight speed according to the performance characteristics of the drone and the requirements of the inspection task. Set a safe flight altitude according to the height and width of the bridge steel structure and the viewing angle and focal length of the drone camera.

[0021] 3. Path planning: Identify obstacles such as bridge support structures and cables that may affect the flight of the drone in the 3D model. Use the Dijkstra path planning algorithm, combined with the flight parameters of the drone and the distribution of obstacles, to calculate the optimal flight path. The specific steps include:

[0022] 3.1. The flight area of ​​the drone is represented as a directed graph, where nodes represent possible flight locations and edges represent the flight path from one location to another.

[0023] 3.2. Collect node information: For each node, collect information about its surrounding environment, including the location of obstacles, terrain features, etc.

[0024] 3.3. Define cost function: Define a cost function for each edge, taking into account factors such as flight distance, risk assessment, and estimated flight time: c i =w1*d i +w2*r i +w3*t i ; Among them: ci is the cost from point i-1 to point i. w1, w2 and w3 are weight coefficients used to balance the importance of different factors. i is the straight-line distance or actual flight distance between two nodes. i is a risk assessment value, which may be based on factors such as the proximity of obstacles, terrain complexity, etc. i is the estimated flight time.

[0025] 3.4. Initialize Dijkstra's algorithm: Select the starting node and set its cost to 0. Set the cost of all other nodes to infinity (indicating that they have not been explored yet).

[0026] 3.5. Explore neighbor nodes: For the current node, explore all its neighbor nodes and calculate the total cost from the starting node to these neighbor nodes.

[0027] 3.6. Update node cost: If the cost of reaching a neighbor node through the current node is less than the cost currently recorded for the neighbor node, then update the cost of the neighbor node.

[0028] 3.7. Select the next node: Select the node with the smallest cost from the unexplored nodes as the next "current node".

[0029] 3.8. Repeat exploration and update: Repeat steps 3.5 to 3.7 until all nodes are explored or the target node is found.

[0030] 3.9. Construct the shortest path: Starting from the target node, trace back to the starting node, and construct the shortest path by tracking the predecessor node of each node.

[0031] 4. Real-time adjustment: Dynamically adjust the flight route according to the actual situation on site and drone status feedback.

[0032] 5. Verification and optimization: Before the formal test, a simulated flight test is conducted to verify the feasibility and safety of the planned path. The path planning algorithm is optimized and adjusted based on the feedback data to improve the detection efficiency and safety.

[0033] Through the above technical solutions, the route planning module can ensure that the UAV can complete the inspection task efficiently and safely in the welding quality inspection of bridge steel structures.

[0034] As an optional solution of the present invention, the image processing and recognition module preprocesses, extracts features and recognizes the images collected by the high-definition camera, and timely identifies the welding quality problems of the bridge steel structure, including the following steps:

[0035] 1. Image preprocessing: including denoising, contrast enhancement and edge detection;

[0036] Denoising: Use algorithms such as Gaussian filtering to smooth the image and reduce noise interference.

[0037] Enhance contrast: Enhance the contrast of the image through techniques such as histogram equalization and contrast stretching. Improve the contrast between different areas in the image to make welding defects (such as cracks, pores, etc.) more clearly visible.

[0038] Edge detection: Use the Sobel operator algorithm to detect edges and obtain clear weld edge images. Detect edge information in the image, especially the edges of the welds.

[0039] 2. Feature extraction: Extract features from the preprocessed image, including color, texture, and shape; use deep learning methods to automatically extract feature information of the welding area, such as crack morphology, pore distribution, etc. Use the residual network (ResNet) deep learning model to automatically learn high-level feature representation of the image.

[0040] 3. Feature selection: Use principal component analysis (PCA) to reduce and select features. Select the most useful features for identifying welding quality problems from the many extracted features to improve the recognition accuracy and efficiency. Calculate the principal component according to the following formula: j =∑ n i＝1 w ij x i ; Among them, z j is the jth principal component, which is a new representation of the original data. ij is the weight of the jth principal component, representing the original feature x i The contribution degree of the jth principal component. i is the i-th feature of the original data. In a multidimensional dataset, each data point can have multiple features, x i represents one of these features. n is the total number of features in the original dataset.

[0041] 4. Recognition: A classification model is constructed using a recurrent neural network (RNN) deep learning model. It is able to classify and identify welding quality problems based on the extracted features. The classification model is trained using a labeled welding quality image dataset. The recognition accuracy and generalization ability of the model are improved by adjusting model parameters and optimizing algorithms. RNN updates its hidden state through time steps, and the formula is expressed as:

[0042] h t =tanh(W hh h t-1 +W xh x t +b h );y t=W hy h t +b y ; In the formula, h t is the hidden state at time step t, which is a vector representing the internal memory or activation value of the RNN at the current time step. tanh is the hyperbolic tangent activation function, which is used to introduce nonlinearity and ensure that the value of the hidden state is between -1 and 1, which helps to handle the complexity of the data. hh is the weight matrix from hidden layer to hidden layer, indicating how the current hidden state depends on the hidden state of the previous time step. t-1 is the hidden state at time step t-1, representing the network’s memory at the previous time step. xh is the weight matrix from the input layer to the hidden layer, indicating how the input of the current time step affects the current hidden state. t is the input vector at time step t. b h is the bias vector for the hidden layer, which adds a constant offset to the calculation of the hidden state and helps to center the activation values.

[0043] The pre-processed images are input into the trained classification model, which will automatically identify welding quality problems and mark the specific location and type. At the same time, the corresponding inspection report or evaluation report can be output based on the recognition results.

[0044] 5. Size comparison: Use the mapping relationship between pixel size and actual size in image processing technology to measure the size of the steel in the image. Compare the measured actual size with the standard size and calculate the size deviation. If the deviation exceeds the allowable range, it is judged as unqualified.

[0045] Through the above technical solutions, the image processing and recognition module can realize timely and accurate identification of welding quality problems of bridge steel structures, providing strong support for subsequent maintenance and repair work.

[0046] As an optional solution of the present invention, the infrared data analysis and identification module processes and analyzes the infrared wave data collected by the infrared thermal imager to identify the welding quality problems of the bridge steel structure; the steps include:

[0047] 1. Data preprocessing: import the infrared image data collected by the infrared thermal imager into the infrared data analysis and recognition module; preprocess the infrared data, including image correction and noise removal;

[0048] 2. Image analysis: Use infrared image data to calculate the temperature distribution of the welded parts of the steel structure of the bridge under test, and identify potential defect areas through temperature differences.

[0049] 3. Defect feature extraction: including crack identification, unfusion identification, pore inclusion identification and deformation identification;

[0050] Crack identification: Observe the infrared image to see if there are any abnormal temperature areas, which may correspond to welding cracks. Cracks usually appear as linear or irregular temperature gradient changes.

[0051] Lack of fusion identification: Lack of fusion areas may appear as uneven temperature distribution in infrared images, with significant differences in temperature from the surrounding normal welding areas.

[0052] Porosity and slag inclusion identification: These defects may appear in infrared images as small, discrete temperature anomalies or areas, which are usually lower or higher than the surrounding material.

[0053] Deformation identification: By comparing the temperature distribution in different areas, it is possible to indirectly determine whether the welding part has deformed. The deformed area may be accompanied by abnormal changes in temperature distribution.

[0054] 4. Quantitative analysis: Use professional image processing and analysis software to conduct quantitative analysis on the extracted defect features, including the size, shape, location and other parameters of the defects.

[0055] 5. Result evaluation: Based on the analysis results, evaluate the defects in the welded parts of the bridge steel structure to determine their severity and impact on the structural safety. Organize the analysis results into a report, including the test object, test method, test results, analysis and evaluation, etc.

[0056] As an optional solution of the present invention, the image analysis comprises the following steps:

[0057] 1. Temperature calibration: Read temperature information from infrared image files; convert pixel values ​​in the image into actual temperature values ​​and create a temperature map.

[0058] 2. Temperature distribution calculation:

[0059] Temperature Histogram: Generates a histogram of temperature distribution to understand the overall temperature range and its distribution.

[0060] Temperature matrix: Divide the image into smaller areas (such as pixel blocks or grids), calculate the average temperature of each area, and form a temperature matrix.

[0061] Temperature Gradient: Calculates the temperature gradient between adjacent regions in a temperature matrix to identify areas of temperature variation.

[0062] Local temperature change analysis of the temperature matrix: Local temperature change is performed using the following formula:

[0063] in, It is the gradient vector of the temperature matrix M in region i, j, which contains the temperature change rate of the point in the x and y directions. It is the partial derivative of the temperature matrix M in the x direction, representing the rate of change of the temperature of region i, j along the x-axis. It is the partial derivative of the temperature matrix M in the y direction, representing the rate of change of the temperature of region i, j along the y axis.

[0064] The temperature gradient is calculated by the following formula:

[0065]

[0066] Among them, G x is the temperature gradient in the x-direction, expressed as the weighted average of the temperature variation along the x-axis. y is the temperature gradient in the y direction, expressed as the weighted average of the temperature change along the y axis. x,y ) and T(x-δ x,y ) represent the displacement δ in the positive and negative directions of the x-axis respectively x The temperature value at T(x,y+δ y ) and T(x,y-δ y ) represent the displacement δ in the positive and negative directions of the y-axis respectively y The temperature value at x ) and w(δ y ) are distance-weighted functions in the x and y directions, respectively. Gaussian functions are used to calculate the gradient for different displacements δ x and δ y Assign different weights to reduce the impact of noise.

[0067] 3. Identification of potential defect areas: including threshold setting, anomaly detection, defect feature analysis, texture feature calculation and defect identification;

[0068] Threshold setting: Set the temperature difference threshold according to the normal temperature range of the bridge steel structure and the welding process requirements.

[0069] Anomaly Detection: By comparing the actual temperature distribution with the normal temperature range, areas with abnormal temperatures are identified. These areas may contain potential defects.

[0070] Defect feature analysis:

[0071] Cracks: Usually appear as linear temperature gradients or areas of temperature discontinuity.

[0072] Porosity and slag inclusions: appear as small, discrete temperature anomalies or areas, where the temperature is lower or higher than the surrounding material.

[0073] Lack of fusion: It is manifested by uneven temperature distribution, which is significantly different from the temperature of the surrounding normal welding area. It includes the following:

[0074] Analysis of temperature distribution inhomogeneity: The lack of fusion area shows a significant difference in temperature distribution from the surrounding normal welding area. This difference can be quantified by calculating the temperature variance of the local area:

[0075]

[0076] Among them, TemV(Ω ij ) represents the local area Ω ij Temperature variance. Ω ij Represents the local area being analyzed, which can be a pixel block or grid in the image. T(p) represents the temperature of pixel p. |Ω ij | represents the local area Ω ij The total number of pixels in μΩ ij Represents area Ω ij The average temperature.

[0077] Calculation of texture features: Use energy and entropy to describe the texture features of the local area, and combine these features through weighted sum:

[0078] UnfusedTexture(Ω ij )=α*Energy(Ω ij )+(1-α)*Entropy(Ω ij );That

[0079] In UnfusedTexture(Ω ij ) represents the local area Ω ij The texture features are synthesized by the weighted sum of energy and entropy. α is a weight coefficient used to balance energy and entropy, and its value is between 0 and 1. ij ) represents the local area Ω ij The texture energy is related to the degree of change of pixel values ​​in the region. ij ) represents the local area Ω ij The texture entropy is related to the randomness of the distribution of pixel values ​​in the area.

[0080] Defect identification: By setting the threshold T unfused To identify the unfused area, the threshold can be calculated based on the temperature variance and texture characteristics of the local area: UnfusedArea = {Ω ij ∣

[0081] UnfusedTexture(Ω ij)>T unfused}; where T unfused It is a preset threshold used to determine whether the texture feature exceeds the normal range, thereby identifying the unfused area. UnfusedArea is the area identified as unfused.

[0082] As an optional solution of the present invention, the comprehensive evaluation module combines the analysis results of the image processing recognition module and the infrared data analysis recognition module to comprehensively evaluate the overall welding quality of the steel structure.

[0083] The following steps are involved:

[0084] 1. Data fusion: The data from the image processing recognition module and the infrared data analysis recognition module are integrated to ensure accurate correspondence and complementarity of defect information. A comprehensive database of welding defects is established, including detailed information of the defects.

[0085] 2. Comprehensive analysis: Evaluate the impact of defects on the overall welding quality of the steel structure based on factors such as the type, location, and severity of the defects. Evaluate the impact of welding defects on the overall safety of the bridge, taking into account factors such as the structural characteristics, load conditions, and use environment of the bridge. The impact assessment is performed through a multi-criteria decision analysis using the following formula:

[0086] ImpactScore i =w ti *P ti +w xi *D xi +w yi *D yi +w si *S si +w ΔTi *D ΔTi ;

[0087] Among them, ImpactScore i is the overall impact score of the ith defect. ti is the weight of defect type ti, reflecting the influence of defect type on the overall score. ti is the severity score of the defect type, which is a quantitative assessment of the impact of the defect type on the structural safety. xi is the weight of the defect position in the x-axis direction. xi is the position-dependency score of the defect in the x-direction, which may be associated with a specific area of ​​the structure (such as a stress concentration area). yi is the weight of the defect position in the y-axis direction. yi is the position-dependent score of the defect in the y-axis direction, and is also related to a specific area of ​​the structure. si is the weight of the defect size si. siIt is the severity score of the defect size, reflecting the impact of the defect size on the structural integrity. ΔTi is the weight of the temperature deviation ΔTi. ΔTi It is the severity score of temperature deviation, reflecting the possible impact of the temperature difference between the welding area and the normal area on the structure.

[0088] 3. Comprehensive evaluation: including quantitative evaluation, qualitative evaluation and comprehensive evaluation;

[0089] Quantitative assessment: Using mathematical models and algorithms, the severity of defects is quantitatively scored. Quantitative assessment is achieved by assigning a standardized score to each influencing factor:

[0090] QS i =a i *ImS i +b i *StF i ; StF i =∑ M j＝1 c ij *Factor j ;

[0091] Among them, QS i is the quantitative evaluation score of the i-th defect. i is the impact score ImS of the i-th defect i The relevant weight coefficient reflects the importance of the defect to the overall structural safety. i is the impact score of the ith defect, which is a quantitative value based on the severity of the defect and its impact on the structural function. i is the structural factor score StF associated with the i-th defect i The relevant weight coefficient is used to adjust the contribution of structural factors to the quantitative evaluation score. i is the comprehensive score of the structural factors of the i-th defect, taking into account the influence of multiple dimensions such as location, size, and load. M is the structural factor score StF i The total number of structural factors considered in c ij It is the weight coefficient of the jth structural factor corresponding to the i-th defect, which is used to weight the impact of each structural factor on the final score. j is the score of the jth structural factor, which is a quantitative value based on a specific standard or scale, reflecting the importance or influence of the factor in the structural evaluation.

[0092] Qualitative assessment: Describe and judge the nature and impact of defects based on expert experience and engineering practice. Evaluate the impact of defects on the local strength and overall stability of the steel structure, consider whether the defect location is in the stress concentration area, and the expansion trend of the defect. Based on the nature and location of the defect, evaluate its potential threat to the overall safety of the bridge, including short-term and long-term impacts. Predict the degree of shortening of the service life of the steel structure by defects, and consider whether it is necessary to repair or replace it in advance.

[0093] Comprehensive evaluation: Combine quantitative and qualitative evaluations to form a comprehensive evaluation of welding quality. Combine expert opinions with quantitative scoring, comprehensively consider the nature, location, severity of defects and their impact on structural performance and safety, and form a more comprehensive and accurate evaluation result.

[0094] Integrate the results of quantitative and qualitative evaluations into a comprehensive evaluation report. The report should include evaluation methods, evaluation criteria, quantitative scoring results, expert opinions, comprehensive judgments, and questions and suggestions.

[0095] 4. Evaluation report: Summarize the evaluation results and write a detailed evaluation report. The report includes evaluation methods, evaluation standards, evaluation results, problems and suggestions, etc.

[0096] The present invention provides a method for detecting the welding quality of a bridge steel structure, comprising the following steps:

[0097] S1. The route planning module uses GIS and CAD technology to build a three-dimensional model of the bridge steel structure based on the design drawings and accurately simulates its spatial layout. Considering the parameters of the drone and the obstacle distribution in the three-dimensional model, the algorithm calculates the optimal flight path to ensure that the drone can complete the inspection task efficiently and safely.

[0098] S2, cruise monitoring module: According to the planned route, the drone carries a high-definition camera and an infrared thermal imager to conduct a full range of welding quality inspections on the bridge;

[0099] S3, the image processing and recognition module pre-processes, extracts features and recognizes the images collected by the high-definition camera, and promptly identifies the welding quality problems of the bridge steel structure, including cracks, lack of fusion, pores, slag inclusions and deformation, etc.; identifies the size of the steel, compares and analyzes it with the specified size, and promptly identifies abnormal situations where the material size does not meet the requirements;

[0100] S4, infrared data analysis and identification module processes and analyzes the infrared wave data collected by the infrared thermal imager to identify the welding quality problems of the bridge steel structure, including cracks, lack of fusion, pores, slag inclusions and deformation;

[0101] S5. The comprehensive evaluation module combines the analysis results of the image processing recognition module and the infrared data analysis recognition module to comprehensively evaluate the overall welding quality of the steel structure. Based on factors such as defect type, location, and severity, the impact of welding quality on the overall safety of the bridge is evaluated.

[0102] S6. When welding quality problems are detected, the alarm module will sound an alarm in time.

[0103] 3. Beneficial effects

[0104] One or more technical solutions provided in the technical solution of this application have at least the following technical effects or advantages:

[0105] 1. The three-dimensional model constructed by the present invention through GIS and CAD technology, combined with the autonomous cruising of drones, can realize efficient and comprehensive detection of bridge steel structures, avoiding omissions and subjective judgment errors that may exist in traditional manual detection, and greatly improving the accuracy and efficiency of detection.

[0106] 2. The present invention can process the collected data in real time through image processing and infrared data analysis modules, and quickly identify welding quality problems, so that problems can be discovered and reported in time, providing a valuable time window for subsequent repair and maintenance work.

[0107] 3. The present invention can perform multi-dimensional detection and comprehensive analysis: the multi-dimensional detection method combined with high-definition cameras and infrared thermal imagers can more comprehensively capture the welding quality information of bridge steel structures, and integrate and analyze various data through a comprehensive evaluation module, thereby improving the comprehensiveness and accuracy of the evaluation.

[0108] 4. The present invention can improve the overall safety of the bridge: by timely discovering and repairing welding quality problems, it can effectively reduce the safety hazards of the bridge structure caused by welding defects and improve the overall safety and service life of the bridge. BRIEF DESCRIPTION OF THE DRAWINGS

[0109] Figure 1 A schematic diagram of a flow chart of a bridge steel structure welding quality inspection method disclosed in a preferred embodiment of the present application;

[0110] Figure 2 This is a schematic structural diagram of a bridge steel structure welding quality inspection device disclosed in a preferred embodiment of the present application. DETAILED DESCRIPTION

[0111] The present application is further described in detail below in conjunction with the accompanying drawings.

[0112] Reference Figure 1 and Figure 2The present application provides a bridge steel structure welding quality detection device, comprising:

[0113] Data acquisition module: collects the design drawings of the bridge steel structure, including the material, specifications and welding requirements (welding position, welding method and weld thickness, etc.) of the steel structure, and collects images of welding defects; annotates the data as a reference sample;

[0114] Route planning module: Based on the design drawings, GIS (geographic information system) and CAD (computer-aided design) technologies are used to build a three-dimensional model of the bridge steel structure and accurately simulate its spatial layout. Considering factors such as the flight speed, altitude, stability, and endurance of the drone, combined with the distribution of obstacles in the three-dimensional model (such as bridge support structures, cables, etc.), the algorithm calculates the optimal flight path to ensure that the drone can complete the inspection task efficiently and safely. During the flight, the flight route is dynamically adjusted according to the actual situation on site (such as wind force, light changes) and drone status feedback to ensure the inspection effect.

[0115] Cruise monitoring module: including drones, infrared thermal imagers, high-definition cameras and LED lights. According to the planned route, drones carry high-definition cameras and infrared thermal imagers to conduct all-round welding quality inspections on bridges; the high-definition camera is equipped with a high-resolution, wide-angle lens and supports night infrared shooting to ensure that welding details can be clearly captured under different lighting conditions. The infrared thermal imager uses the characteristics of infrared thermal imagers that can capture and record the temperature distribution on the surface of objects, thereby discovering possible internal defects or quality problems in the welding parts. Detect internal defects in welding, such as cracks, lack of fusion, etc.

[0116] Image processing and recognition module: pre-processes, extracts features and recognizes images collected by high-definition cameras, and promptly identifies welding quality problems of bridge steel structures, including cracks, lack of fusion, pores, slag inclusions and deformation; identifies the size of steel materials, compares and analyzes them with the specified size, and promptly identifies abnormal situations where the material size does not meet the requirements;

[0117] Infrared data analysis and identification module: Process and analyze the infrared wave data collected by the infrared thermal imager to identify welding quality problems of bridge steel structures, including cracks, lack of fusion, pores, slag inclusions and deformation;

[0118] Comprehensive evaluation module: Combines the analysis results of the image processing recognition module and the infrared data analysis recognition module to comprehensively evaluate the overall welding quality of the steel structure. Based on factors such as defect type, location, and severity, the impact of welding quality on the overall safety of the bridge is evaluated.

[0119] Alarm module: includes an alarm, which will sound an alarm in time when welding quality problems are detected; according to the evaluation results, different levels of alarm thresholds are set, such as minor defect reminders, serious defect warnings, etc. Alarm information is sent to the control center and related personnel in real time through the wireless network to ensure that the problem is handled in time.

[0120] Control center: connected to the data collection module, route planning module, cruise monitoring module, image processing and recognition module, ultrasonic data analysis and recognition module, comprehensive evaluation module and alarm module. Receive and display the data and status of each module in real time, support remote control of drones and detection equipment. Provide powerful data analysis tools, support historical data query, trend analysis, report generation and other functions.

[0121] Furthermore, the route planning module plans the flight route of the drone, including the following steps:

[0122] 1. Build a 3D model: First, take the design drawings of the bridge steel structure as input. These drawings should contain detailed information such as the size, shape, location, and welding requirements of the steel structure. Based on the design drawings, use geographic information system (GIS) and computer-aided design (CAD) technology to build a 3D model of the bridge steel structure. GIS technology is used to process geospatial data, while CAD technology excels in accurate drawing and modeling. Through 3D modeling, the spatial layout of the bridge steel structure is accurately simulated, including key parts such as the bridge's supporting structure, cables, and welded joints.

[0123] 2. Set the flight parameters of the drone: Set the appropriate flight speed according to the performance characteristics of the drone and the requirements of the inspection task. The flight speed should ensure the inspection efficiency while avoiding image blur or data loss caused by too fast a speed. Set a safe flight altitude according to the height and width of the bridge steel structure and the viewing angle and focal length of the drone camera. The flight altitude should ensure that the drone can clearly capture the details of the welding area while avoiding collision with the bridge structure. Consider the flight stability of the drone to ensure that it can maintain stable flight under the interference of external factors such as wind and airflow.

[0124] Evaluate the drone’s endurance based on its battery capacity and the estimated duration of the inspection mission to ensure it can complete the entire inspection mission without having to replace the battery midway.

[0125] 3. Path planning: Identify obstacles such as bridge support structures and cables that may affect the flight of the drone in the 3D model. Use the Dijkstra path planning algorithm, combined with the flight parameters of the drone and the distribution of obstacles, to calculate the optimal flight path. The optimal path should be as short as possible while avoiding all obstacles to ensure that the drone can fly safely and efficiently. The specific steps include:

[0126] 3.1. The flight area of ​​the drone is represented as a directed graph, where nodes represent possible flight locations and edges represent the flight path from one location to another.

[0127] 3.2. Collect node information: For each node, collect information about its surrounding environment, including the location of obstacles, terrain features, etc.

[0128] 3.3. Define cost function: Define a cost function for each edge, which takes into account factors such as flight distance, risk assessment and estimated flight time: c i =w1*d i +w2*r i +w3*t i ; Among them: c i is the cost from point i-1 to point i. w1, w2 and w3 are weight coefficients used to balance the importance of different factors. i is the straight-line distance or actual flight distance between two nodes. i is a risk assessment value, which may be based on factors such as the proximity of obstacles, terrain complexity, etc. i is the estimated flight time, which may be based on factors such as flight distance, speed limitations and environmental conditions.

[0129] 3.4. Initialize Dijkstra's algorithm: Select the starting node and set its cost to 0. Set the cost of all other nodes to infinity (indicating that they have not been explored yet).

[0130] 3.5. Explore neighbor nodes: For the current node, explore all its neighbor nodes and calculate the total cost from the starting node to these neighbor nodes.

[0131] 3.6. Update node cost: If the cost of reaching a neighbor node through the current node is less than the cost currently recorded for the neighbor node, then update the cost of the neighbor node.

[0132] 3.7. Select the next node: Select the node with the smallest cost from the unexplored nodes as the next "current node".

[0133] 3.8. Repeat exploration and update: Repeat steps 3.5 to 3.7 until all nodes are explored or the target node is found.

[0134] 3.9. Build the shortest path: Start from the target node and trace back to the starting node, building the shortest path by tracking the predecessor nodes of each node. Smooth the generated path to avoid sharp turns or unnatural path changes. Make sure all points on the path have enough safety distance to avoid obstacles.

[0135] 4. Real-time adjustment: During the flight of the drone, the actual conditions of wind, light, etc. on site are monitored in real time through the sensors and cameras carried by the drone. The status information sent by the drone (such as battery power, flight speed, altitude, etc.) is received to ensure that the drone is in normal working condition. The flight route is dynamically adjusted according to the actual situation on site and the drone status feedback. For example, the flight altitude and speed can be reduced when the wind suddenly increases; the LED light can be turned on for auxiliary lighting when the light is insufficient.

[0136] 5. Verification and optimization: Before the formal test, a simulated flight test is conducted to verify the feasibility and safety of the planned path. During the test, feedback data such as the flight data and captured images of the drone are collected. The path planning algorithm is optimized and adjusted based on the feedback data to improve the detection efficiency and safety.

[0137] In this technical solution, the route planning module can ensure that the drone can complete the inspection task efficiently and safely in the welding quality inspection of bridge steel structures.

[0138] Furthermore, the image processing and recognition module pre-processes, extracts features and recognizes the images collected by the high-definition camera, and timely identifies the welding quality problems of the bridge steel structure, including the following steps:

[0139] 1. Image preprocessing: including denoising, contrast enhancement, edge detection, etc.; image preprocessing is a key step to improve image quality, reduce noise and interference, and lay a good foundation for subsequent feature extraction and recognition.

[0140] Denoising: Remove noise from the image, such as random noise, Gaussian noise, etc., which will interfere with subsequent image processing and analysis. Use algorithms such as median filtering and Gaussian filtering to smooth the image and reduce noise interference.

[0141] Enhance contrast: Improve the contrast between different areas in the image to make welding defects (such as cracks, pores, etc.) more visible. Enhance the contrast of the image through techniques such as histogram equalization and contrast stretching.

[0142] Edge detection: Detect edge information in the image, especially the edge of the weld, which is crucial for subsequent feature extraction and recognition. Use algorithms such as the Sobel operator and Canny edge detector to perform edge detection and obtain a clear weld edge image.

[0143] 2. Feature extraction: Extract features from preprocessed images, including color, texture, and shape; use deep learning methods to automatically extract feature information of the welding area, such as crack morphology, pore distribution, etc. Use the residual network (ResNet) deep learning model to automatically learn high-level feature representations of images. These models can automatically learn from large amounts of data and extract the most useful features for identifying welding quality problems.

[0144] 3. Feature selection: Use principal component analysis (PCA) to reduce and select features. Select the most useful features for identifying welding quality problems from the many extracted features to improve the recognition accuracy and efficiency. Calculate the principal component according to the following formula: j =∑ n i＝1 w ij x i ; Among them, z j is the jth principal component. It is a new representation of the original data, which is the projection of the original data in the direction of the principal component. ij is the weight of the jth principal component, also called loading or coefficient. It represents the original feature x i The contribution of x to the jth principal component. These weights form the jth row of the transformation matrix. i is the i-th feature of the original data. In a multidimensional dataset, each data point can have multiple features, x i represents one of these features. n is the total number of features in the original dataset.

[0145] 4. Recognition: A classification model is constructed using a recurrent neural network (RNN) deep learning model. It is able to classify and identify welding quality problems based on the extracted features. The classification model is trained using a labeled welding quality image dataset. The recognition accuracy and generalization ability of the model are improved by adjusting model parameters and optimizing algorithms. RNN updates its hidden state through time steps, and the formula is expressed as:

[0146] h t =tanh(W hh h t-1 +W xh x t +b h );y t =W hy h t +b y ; In the formula, h t is the hidden state at time step t, which is a vector representing the internal memory or activation value of the RNN at the current time step. tanh is the hyperbolic tangent activation function, which is used to introduce nonlinearity and ensure that the value of the hidden state is between -1 and 1, which helps to handle the complexity of the data. hhis the weight matrix from hidden layer to hidden layer, indicating how the current hidden state depends on the hidden state of the previous time step. t-1 is the hidden state at time step t-1, representing the network’s memory at the previous time step. xh is the weight matrix from the input layer to the hidden layer, indicating how the input of the current time step affects the current hidden state. t is the input vector at time step t, which can be a sample in the sequence data. For example, in welding defect detection, it can be an image feature captured at a certain time point. h is the bias vector for the hidden layer, which adds a constant offset to the calculation of the hidden state and helps to center the activation values.

[0147] The pre-processed images are input into the trained classification model, which will automatically identify welding quality problems and mark the specific location and type. At the same time, the corresponding inspection report or evaluation report can be output based on the recognition results.

[0148] 5. Dimension comparison: Dimension comparison is to use image measurement technology to automatically measure the actual size of steel and compare it with the standard size to identify dimensional deviations.

[0149] The mapping relationship between pixel size and actual size in image processing technology is used to measure the size of steel in the image. The actual size measured is compared with the standard size to calculate the size deviation. If the deviation exceeds the allowable range, it is judged as unqualified.

[0150] In this technical solution, the image processing and recognition module can realize timely and accurate identification of welding quality problems of bridge steel structures, providing strong support for subsequent maintenance and repair work.

[0151] Furthermore, the infrared data analysis and recognition module processes and analyzes the infrared wave data collected by the infrared thermal imager to identify the welding quality problems of the bridge steel structure; the following steps are included:

[0152] 1. Data preprocessing: import the infrared image data collected by the infrared thermal imager into the infrared data analysis and recognition module; preprocess the infrared data, including image correction and noise removal;

[0153] Image correction: Correct the imported infrared images, including geometric correction, radiation correction, etc., to eliminate image distortion and errors and ensure data accuracy.

[0154] Noise removal: Remove noise from the image through filtering algorithms (such as median filtering, Gaussian filtering, etc.) to improve the signal-to-noise ratio and clarity of the image.

[0155] 2. Image analysis: Use infrared image data to calculate the temperature distribution of the welded parts of the steel structure of the bridge under test, and identify potential defect areas through temperature differences.

[0156] 3. Defect feature extraction:

[0157] Crack identification: Observe the infrared image to see if there are any abnormal temperature areas, which may correspond to welding cracks. Cracks usually appear as linear or irregular temperature gradient changes.

[0158] Lack of fusion identification: Lack of fusion areas may appear as uneven temperature distribution in infrared images, with significant differences in temperature from the surrounding normal welding areas.

[0159] Porosity and slag inclusion identification: These defects may appear as small, discrete temperature anomalies or areas in infrared images, whose temperatures are usually lower or higher than the surrounding materials.

[0160] Deformation identification: By comparing the temperature distribution in different areas, it is possible to indirectly determine whether the welding part has deformed. The deformed area may be accompanied by abnormal changes in temperature distribution.

[0161] 4. Quantitative analysis: Use professional image processing and analysis software to conduct quantitative analysis on the extracted defect features, including the size, shape, location and other parameters of the defects.

[0162] 5. Result evaluation: Based on the analysis results, evaluate the defects in the welding parts of the bridge steel structure to determine their severity and impact on the structural safety.

[0163] The analysis results shall be compiled into a report, including the test object, test method, test results, analysis and evaluation, etc. The report shall clearly indicate the type, location, size of the defects detected and the recommended treatment measures.

[0164] Further, the image analysis includes the following steps:

[0165] 1. Temperature calibration:

[0166] Read Temperature Information: Reads temperature information from an infrared image file, which is typically associated with each pixel in the image.

[0167] Temperature Mapping: Convert pixel values ​​in an image to actual temperature values ​​and create a temperature map.

[0168] 2. Temperature distribution calculation:

[0169] Temperature Histogram: Generates a histogram of temperature distribution to understand the overall temperature range and its distribution.

[0170] Temperature matrix: Divide the image into smaller areas (such as pixel blocks or grids), calculate the average temperature of each area, and form a temperature matrix.

[0171] Temperature Gradient: Calculates the temperature gradient between adjacent regions in a temperature matrix to identify areas of temperature variation.

[0172] Local temperature change analysis of the temperature matrix: Local temperature change is performed using the following formula:

[0173] in, It is the gradient vector of the temperature matrix M in region i, j, which contains the temperature change rate of the point in the x and y directions. It is the partial derivative of the temperature matrix M in the x direction, representing the rate of change of the temperature of region i, j along the x-axis. It is the partial derivative of the temperature matrix M in the y direction, representing the rate of change of the temperature of region i, j along the y axis.

[0174] The temperature gradient is calculated by the following formula:

[0175]

[0176] Among them, G x is the temperature gradient in the x-direction, expressed as the weighted average of the temperature variation along the x-axis. y is the temperature gradient in the y direction, expressed as the weighted average of the temperature change along the y axis. x,y ) and T(x-δ x,y ) represent the displacement δ in the positive and negative directions of the x-axis respectively x The temperature value at T(x,y+δ y ) and T(x,y-δ y ) represent the displacement δ in the positive and negative directions of the y-axis respectively y The temperature value at x ) and w(δ y ) are distance-weighted functions in the x and y directions, respectively. Gaussian functions are used to calculate the gradient for different displacements δ x and δ y Assign different weights to reduce the impact of noise.

[0177] 3. Identification of potential defect areas:

[0178] Threshold setting: Set the temperature difference threshold according to the normal temperature range of the bridge steel structure and the welding process requirements.

[0179] Anomaly Detection: By comparing the actual temperature distribution with the normal temperature range, areas with abnormal temperatures are identified. These areas may contain potential defects.

[0180] Defect feature analysis:

[0181] Cracks: Usually appear as linear temperature gradients or areas of temperature discontinuity.

[0182] Porosity and slag inclusions: appear as small, discrete temperature anomalies or areas, where the temperature is lower or higher than the surrounding material.

[0183] Lack of fusion: It is manifested by uneven temperature distribution, which is significantly different from the temperature of the surrounding normal welding area. It includes the following:

[0184] Analysis of temperature distribution inhomogeneity: The lack of fusion area shows a significant difference in temperature distribution from the surrounding normal welding area. This difference can be quantified by calculating the temperature variance of the local area:

[0185]

[0186] Among them, TemV(Ω ij ) represents the local area Ω ij Temperature variance. Ω ij Represents the local area being analyzed, which can be a pixel block or grid in the image. T(p) represents the temperature of pixel p. |Ω ij | represents the local areaΩ ij The total number of pixels in μΩ ij Represents area Ω ij The average temperature.

[0187] Calculation of texture features: Use energy and entropy to describe the texture features of the local area, and combine these features through weighted sum:

[0188] UnfusedTexture(Ω ij )=α*Energy(Ω ij )+(1-α)*Entropy(Ω ij );That

[0189] In UnfusedTexture(Ω ij ) represents the local area Ω ij The texture features are synthesized by the weighted sum of energy and entropy. α is a weight coefficient used to balance energy and entropy, and its value is between 0 and 1. ij ) represents the local area Ω ij The texture energy is related to the degree of change of pixel values ​​in the region. ij ) represents the local area Ω ijThe texture entropy is related to the randomness of the distribution of pixel values ​​in the area.

[0190] Defect identification: By setting the threshold T unfused To identify the unfused area, the threshold can be calculated based on the temperature variance and texture characteristics of the local area: UnfusedArea = {Ω ij ∣

[0191] UnfusedTexture(Ω ij )>T unfused}; where T unfused It is a preset threshold used to determine whether the texture feature exceeds the normal range, thereby identifying the unfused area. UnfusedArea is the area identified as unfused.

[0192] Furthermore, the comprehensive evaluation module combines the analysis results of the image processing recognition module and the infrared data analysis recognition module to comprehensively evaluate the overall welding quality of the steel structure. The following steps are included:

[0193] 1. Data fusion: The data from the image processing recognition module and the infrared data analysis recognition module are integrated to ensure accurate correspondence and complementarity of defect information. A comprehensive database of welding defects is established, including detailed information of the defects (type, location, size, temperature anomaly, etc.).

[0194] 2. Comprehensive analysis: Evaluate the impact of defects on the overall welding quality of the steel structure based on factors such as the type, location, and severity of the defects. Evaluate the impact of welding defects on the overall safety of the bridge, taking into account factors such as the structural characteristics, load conditions, and use environment of the bridge. The impact assessment is performed through a multi-criteria decision analysis using the following formula:

[0195] ImpactScore i =w ti *P ti +w xi *D xi +w yi *D yi +w si *S si +w ΔTi *D ΔTi ;

[0196] Among them, ImpactScore i is the overall impact score of the ith defect. ti is the weight of defect type ti, reflecting the influence of defect type on the overall score. ti is the severity score of the defect type, which is a quantitative assessment of the impact of the defect type on the structural safety. xiis the weight of the defect position in the x-axis direction. xi is the position-dependency score of the defect in the x-direction, which may be associated with a specific area of ​​the structure (such as a stress concentration area). yi is the weight of the defect position in the y-axis direction. yi is the position-dependent score of the defect in the y-axis direction, and is also related to a specific area of ​​the structure. si is the weight of the defect size si. si It is the severity score of the defect size, reflecting the impact of the defect size on the structural integrity. ΔTi is the weight of the temperature deviation ΔTi. ΔTi It is the severity score of temperature deviation, reflecting the possible impact of the temperature difference between the welding area and the normal area on the structure.

[0197] 3. Comprehensive evaluation:

[0198] Quantitative assessment: Use mathematical models and algorithms to quantify the severity of defects.

[0199] Quantitative evaluation is achieved by assigning a standardized score to each influencing factor:

[0200] QS i =a i *ImS i +b i *StF i ; StF i =∑ M j＝1 c ij *Factor j ;

[0201] Among them, QS i is the quantitative evaluation score of the i-th defect. i is the impact score ImS of the i-th defect i The relevant weight coefficient reflects the importance of the defect to the overall structural safety. i is the impact score of the ith defect, which is a quantitative value based on the severity of the defect and its impact on the structural function. i is the structural factor score StF associated with the i-th defect i The relevant weight coefficient is used to adjust the contribution of structural factors to the quantitative evaluation score. i is the comprehensive score of the structural factors of the i-th defect, taking into account the influence of multiple dimensions such as location, size, and load. M is the structural factor score StF i The total number of structural factors considered in c ijIt is the weight coefficient of the jth structural factor corresponding to the i-th defect, which is used to weight the impact of each structural factor on the final score. j is the score of the jth structural factor, which is a quantitative value based on a specific standard or scale, reflecting the importance or influence of the factor in the structural evaluation.

[0202] Qualitative assessment: Describe and judge the nature and impact of defects based on expert experience and engineering practice. Evaluate the impact of defects on the local strength and overall stability of the steel structure, consider whether the defect location is in the stress concentration area, and the expansion trend of the defect. Based on the nature and location of the defect, evaluate its potential threat to the overall safety of the bridge, including short-term and long-term impacts. Predict the degree of shortening of the service life of the steel structure by defects, and consider whether it is necessary to repair or replace it in advance.

[0203] Comprehensive evaluation: Combine quantitative and qualitative evaluations to form a comprehensive evaluation of welding quality. Combine expert opinions with quantitative scoring, comprehensively consider the nature, location, severity of defects and their impact on structural performance and safety, and form a more comprehensive and accurate evaluation result.

[0204] Integrate the results of quantitative and qualitative evaluations into a comprehensive evaluation report. The report should include evaluation methods, evaluation criteria, quantitative scoring results, expert opinions, comprehensive judgments, and questions and suggestions.

[0205] 4. Evaluation report: Summarize the evaluation results and write a detailed evaluation report. The report should include evaluation methods, evaluation standards, evaluation results, problems and suggestions. Feedback the evaluation report to relevant units and personnel, including construction units, supervision units, design units, etc. According to the evaluation results, put forward improvement measures and suggestions, such as strengthening welding process control, improving welder skills, optimizing construction management, etc.

[0206] Reference Figure 1 The present invention provides a method for detecting the welding quality of a bridge steel structure, comprising the following steps:

[0207] S1. The route planning module uses GIS (geographic information system) and CAD (computer-aided design) technology to build a three-dimensional model of the bridge steel structure based on the design drawings and accurately simulates its spatial layout. Considering factors such as the flight speed, altitude, stability, and endurance of the drone, combined with the distribution of obstacles in the three-dimensional model (such as bridge support structures, cables, etc.), the algorithm calculates the optimal flight path to ensure that the drone can complete the inspection task efficiently and safely.

[0208] S2, cruise monitoring module: According to the planned route, the drone carries a high-definition camera and an infrared thermal imager to conduct a full range of welding quality inspections on the bridge;

[0209] S3, the image processing and recognition module pre-processes, extracts features and recognizes the images collected by the high-definition camera, and promptly identifies the welding quality problems of the bridge steel structure, including cracks, lack of fusion, pores, slag inclusions and deformation, etc.; identifies the size of the steel, compares and analyzes it with the specified size, and promptly identifies abnormal situations where the material size does not meet the requirements;

[0210] S4, infrared data analysis and identification module processes and analyzes the infrared wave data collected by the infrared thermal imager to identify the welding quality problems of the bridge steel structure, including cracks, lack of fusion, pores, slag inclusions and deformation;

[0211] S5. The comprehensive evaluation module combines the analysis results of the image processing recognition module and the infrared data analysis recognition module to comprehensively evaluate the overall welding quality of the steel structure. Based on factors such as defect type, location, and severity, the impact of welding quality on the overall safety of the bridge is evaluated.

[0212] S6. When welding quality problems are detected, the alarm module will issue an alarm in time; according to the evaluation results, different levels of alarm thresholds are set, such as minor defect reminders, serious defect warnings, etc.

[0213] The working principle of the bridge steel structure welding quality inspection device of the present invention is as follows: the route planning module uses GIS and CAD technology to build a three-dimensional model of the bridge steel structure based on the design drawings to accurately simulate its spatial layout. Considering the flight parameter factors of the drone and combining the obstacle distribution in the three-dimensional model, the optimal flight path is calculated through algorithm to ensure that the drone can complete the inspection task efficiently and safely. The cruise monitoring module follows the planned route, and the drone carries a high-definition camera and an infrared thermal imager to conduct a comprehensive welding quality inspection of the bridge; the image processing and recognition module preprocesses, extracts features and recognizes the images collected by the high-definition camera, and promptly identifies the welding quality problems of the bridge steel structure, including cracks, lack of fusion, pores, slag inclusions and deformation, etc.; identifies the size of the steel, compares and analyzes it with the specified size, and promptly identifies the abnormal situation that the material size does not meet the requirements; the infrared data analysis and recognition module processes and analyzes the infrared wave data collected by the infrared thermal imager, and identifies the welding quality problems of the bridge steel structure, including cracks, lack of fusion, pores, slag inclusions and deformation, etc.; the comprehensive evaluation module combines the analysis results of the image processing and recognition module and the infrared data analysis and recognition module to conduct a comprehensive evaluation of the overall welding quality of the steel structure. According to factors such as defect type, location, and severity, the impact of welding quality on the overall safety of the bridge is evaluated. When a welding quality problem is detected, the alarm module promptly issues an alarm; according to the evaluation results, different levels of alarm thresholds are set, such as minor defect reminders and serious defect warnings.

[0214] The present invention uses a three-dimensional model constructed by GIS and CAD technology, combined with autonomous cruising of drones, to achieve efficient and comprehensive detection of bridge steel structures, avoiding possible omissions and subjective judgment errors in traditional manual detection, and greatly improving the accuracy and efficiency of detection. The image processing and infrared data analysis modules can process the collected data in real time, and quickly identify welding quality problems, so that problems can be discovered and reported in a timely manner, providing a valuable time window for subsequent repair and maintenance work. The multi-dimensional detection method combined with high-definition cameras and infrared thermal imagers can more comprehensively capture the welding quality information of bridge steel structures, and integrate and analyze various data through a comprehensive evaluation module, thereby improving the comprehensiveness and accuracy of the evaluation.

[0215] As described above, the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features thereof may be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for detecting the welding quality of a bridge steel structure, characterized in that: The following steps are involved: S1. The route planning module uses GIS and CAD technology to build a three-dimensional model of the bridge steel structure based on the design drawings, taking into account the parameters of the drone and the obstacle distribution in the three-dimensional model, and calculates the optimal flight path through algorithms; S2, cruise monitoring module: According to the planned route, the drone carries a high-definition camera and an infrared thermal imager to conduct a full range of welding quality inspections on the bridge; S3, the image processing and recognition module pre-processes, extracts features and recognizes the images collected by the high-definition camera, identifies the welding quality problems of the bridge steel structure, and identifies the abnormal situation that the material size does not meet the requirements; S4, the infrared data analysis and recognition module processes and analyzes the infrared wave data collected by the infrared thermal imager to identify the welding quality problems of the bridge steel structure; S5, the comprehensive evaluation module combines the analysis results of the image processing recognition module and the infrared data analysis recognition module to comprehensively evaluate the overall welding quality of the steel structure; based on the defect type, location, and severity factors, it evaluates the impact of welding quality on the overall safety of the bridge; S6. When welding quality problems are detected, the alarm module will sound an alarm in time.

2. The bridge steel structure welding quality inspection method according to claim 1 is characterized in that: Step S1 includes the following steps: S11. Constructing a 3D model: Based on the design drawings of the bridge steel structure, using the Geographic Information System (GIS) and Computer-Aided Design (CAD) technology, construct a 3D model of the bridge steel structure. Through 3D modeling, accurately simulate the spatial layout of the bridge steel structure, including the bridge's supporting structure, cables, and key welded joints. S12. Setting the flight parameters of the UAV: ​​setting the appropriate flight speed and altitude according to the performance characteristics of the UAV and the requirements of the detection task; S13, Path Planning: Identify obstacles that may affect the flight of the UAV in the 3D model; use the Dijkstra path planning algorithm, combined with the flight parameters of the UAV and the distribution of obstacles, to calculate the optimal flight path; S14, Real-time adjustment: Dynamically adjust the flight route according to the actual situation on site and the drone status feedback; S15. Verification and optimization: Before the formal test, a simulated flight test is conducted to verify the feasibility and safety of the planned path; the path planning algorithm is optimized and adjusted based on the feedback data.

3. The bridge steel structure welding quality inspection method according to claim 2 is characterized by: Step S13 includes the following steps: S13.

1. Represent the flight area of ​​the drone as a directed graph, where nodes represent possible flight locations and edges represent flight paths from one location to another. S13.

2. Collect node information: For each node, collect information about its surrounding environment, including the location of obstacles and terrain characteristics; S13.

3. Define cost function: Define a cost function for each edge: c i =w1*d i +w2*r i +w3*t i ; Among them: c i is the cost from point i-1 to point i; w1, w2 and w3 are weight coefficients; d i is the straight-line distance or actual flight distance between two nodes; r i is the risk assessment value; t i is the estimated flight time; S13.

4. Initialize Dijkstra algorithm: select the starting node and set its cost to 0, and set the costs of all other nodes to infinity; S13.5, explore neighbor nodes: for the current node, explore all its neighbor nodes and calculate the total cost from the start node to these neighbor nodes; S13.6, Update node cost: If the cost of reaching a neighbor node through the current node is less than the cost currently recorded for the neighbor node, then update the cost of the neighbor node; S13.7, select the next node: select the node with the minimum cost from the unexplored nodes as the next "current node"; S13.8, Repeat exploration and update: Repeat steps S13.5 to S13.7 until all nodes are explored or the target node is found; S13.

9. Build the shortest path: Starting from the target node, trace back to the starting node, and build the shortest path by tracking the predecessor node of each node.

4. The bridge steel structure welding quality inspection method according to claim 1 is characterized by: Step S3 includes the following steps: S31, Image preprocessing: including denoising, contrast enhancement and edge detection; S32, feature extraction: extracting features from the preprocessed image, the extracted features including color, texture and shape; S33, feature selection: select the most useful features for welding quality problem identification from the extracted features to improve the recognition accuracy and efficiency; calculate the principal component according to the following formula: j =∑ n i＝1 w ij x i ; Among them, z j is the jth principal component; w ij is the weight of the jth principal component, representing the original feature x i The contribution degree of the jth principal component; x i is the i-th feature of the original data; n is the total number of features in the original data set; S34, Recognition: Use the recurrent neural network (RNN) deep learning model to build a classification model; use the labeled welding quality image dataset to train the classification model; improve the recognition accuracy and generalization ability of the model by adjusting the model parameters and optimizing the algorithm; RNN updates its hidden state through time steps, and the formula is expressed as: h t =tanh(W hh h t-1 +W xh x t +b h );y t =W hy h t +b y ; In the formula, h t is the hidden state at time step t, which represents the internal memory or activation value of the RNN at the current time step; tanh is the hyperbolic tangent activation function, which ensures that the value of the hidden state is between -1 and 1; W hh is the weight matrix from hidden layer to hidden layer; h t-1 is the hidden state at time step t-1; W xh is the weight matrix from the input layer to the hidden layer; x t is the input vector at time step t; b h is the bias vector of the hidden layer; the preprocessed image is input into the trained classification model, and the model automatically identifies the welding quality problems and marks the specific location and type; S35, size comparison: Utilize the mapping relationship between pixel size and actual size in image processing technology to measure the size of steel in the image; compare the measured actual size with the standard size and calculate the size deviation; if the deviation exceeds the allowable range, it is judged as unqualified size.

5. The bridge steel structure welding quality inspection method according to claim 1 is characterized by: Step S4 includes the following steps: S41, data preprocessing: importing the infrared image data collected by the infrared thermal imager into the infrared data analysis and recognition module; preprocessing the infrared data, including image correction and noise removal; S42, Image analysis: Use infrared image data to calculate the temperature distribution of the welded parts of the steel structure of the bridge under test, and identify potential defect areas through temperature differences; S43, defect feature extraction: including crack identification, lack of fusion identification, pore inclusion identification and deformation identification; S44, quantitative analysis: Use professional image processing and analysis software to conduct quantitative analysis on the extracted defect features, including the size, shape and location parameters of the defects; S45. Result evaluation: Based on the analysis results, evaluate the defects in the welding parts of the bridge steel structure to determine their severity and impact on the structural safety; organize the analysis results into a report.

6. The bridge steel structure welding quality inspection method according to claim 5 is characterized by: Step S42 includes the following steps: S421, temperature calibration: reading temperature information from the infrared image file; converting pixel values ​​in the image into actual temperature values, and creating a temperature map; S422, temperature distribution calculation: generate a histogram of temperature distribution to understand the overall temperature range and its distribution; divide the image into smaller areas, calculate the average temperature of each area, and form a temperature matrix; calculate the temperature gradient between adjacent areas in the temperature matrix to identify areas with temperature changes; S423, potential defect area identification: including threshold setting, anomaly detection, defect feature analysis, texture feature calculation and defect identification; S4231, threshold setting: set the threshold of temperature difference according to the normal temperature range of bridge steel structure and welding process requirements; S4232, abnormality detection: identifying areas with abnormal temperature by comparing the actual temperature distribution with the normal temperature range; S4233. Analysis of defect characteristics: Cracks appear as linear temperature gradient changes or temperature discontinuity areas; pores and slag inclusions appear as small, discrete temperature anomaly points or areas, whose temperatures are lower or higher than the surrounding materials; lack of fusion appears as uneven temperature distribution, which is significantly different from the temperature of the surrounding normal welding area.

7. The bridge steel structure welding quality inspection method according to claim 6 is characterized by: In step S422, the local temperature change analysis of the temperature matrix is ​​performed by the following formula: in, is the gradient vector of the temperature matrix M at region i, j, which contains the temperature change rate of the point in the x and y directions; represents the rate of change of temperature in region i, j along the x-axis; represents the rate of change of temperature in region i, j along the y-axis; The temperature gradient is calculated by the following formula: Among them, G x is the temperature gradient in the x direction; G y is the temperature gradient in the y direction; T(x+δ x,y ) and T(x-δ x,y ) represent the displacement δ in the positive and negative directions of the x-axis respectively x The temperature value at T(x,y+δ y ) and T(x,y-δ y ) represent the positive and negative displacements of the y-axis, δ y The temperature value at x ) and w(δ y ) are distance-weighted functions in the x and y directions, respectively.

8. The bridge steel structure welding quality inspection method according to claim 6 is characterized by: Step S4233 includes the following steps: S4233.

1. Analysis of temperature distribution inhomogeneity: The temperature distribution of the unfused area is significantly different from that of the surrounding normal welding area, which is quantified by calculating the temperature variance of the local area: Among them, TemV(Ω ij ) represents the local area Ω ij Temperature variance, Ω ij represents the local area being analyzed, T(p) represents the temperature of pixel p, |Ω ij | represents the local area Ω ij Total number of internal pixels, μΩ ij Represents area Ω ij The average temperature of S4233.

2. Calculate texture features: Use energy and entropy to describe the texture features of the local area, and combine these features through weighted sums: UnfusedTexture(Ω ij )=α*Energy(Ω ij )+(1-α)*Entropy(Ω ij ); among them, UnfusedTexture(Ω ij ) represents the local area Ω ij Texture features; α is a weight coefficient used to balance energy and entropy; Energy(Ω ij ) represents the local area Ω ij Texture energy, Entropy(Ω ij ) represents the local area Ω ij The texture entropy of S4233.

3. Defect identification: Identify unfused areas by setting thresholds.

9. The bridge steel structure welding quality inspection method according to claim 1 is characterized by: Step S5 includes the following steps: S51, data fusion: fuse the data of the image processing recognition module and the infrared data analysis recognition module to establish a comprehensive database of welding defects; S52. Comprehensive analysis: Based on the type, location and severity of defects, evaluate their impact on the overall welding quality of the steel structure; the impact assessment is performed using the following formula: ImpactScore i = w ti * P ti + w xi * D xi + w yi * D yi + w si * S si + w ΔTi * D ΔTi ; where, ImpactScore i is the overall impact score of the ith defect, w ti is the weight of defect type ti, P ti is the severity score of the defect type, w xi is the weight of the defect position in the x-axis direction, D xi is the position correlation score of the defect in the x-axis direction, w yi is the weight of the defect position in the y-axis direction, D yi is the position correlation score of the defect in the y-axis direction, w si is the weight of the defect size si, S si is the severity score of the defect size, w ΔTi is the weight of the temperature deviation ΔTi, D ΔTi is the severity score of the temperature deviation; S53. Comprehensive evaluation: including quantitative evaluation, qualitative evaluation and comprehensive evaluation; S531. Quantitative assessment: The severity of defects is quantitatively scored using mathematical models and algorithms; quantitative assessment is achieved by assigning a standardized score to each influencing factor: QS i =a i *ImS i +b i *StF i ; StF i =∑ M j＝1 c ij *Factor j Among them, QS i is the quantitative evaluation score of the i-th defect, a i is the impact score ImS of the i-th defect i The relevant weight coefficient, ImS i is the impact score of the ith defect, b i is the structural factor score StF associated with the i-th defect i The relevant weight coefficient, StF i is the comprehensive score of the structural factors of the i-th defect, and M is the structural factor score StF i The total number of structural factors considered in ij is the weight coefficient of the jth structural factor corresponding to the i-th defect, Factor j is the score of the jth structural factor; S532, Qualitative assessment: Describe and judge the nature and impact of defects by combining expert experience and engineering practice; S533, Comprehensive evaluation: Combine quantitative and qualitative evaluations to form a comprehensive evaluation of welding quality; Integrate the results of quantitative and qualitative evaluations into a comprehensive evaluation report; S54. Evaluation report: Summarize the evaluation results and write a detailed evaluation report.

10. A bridge steel structure welding quality detection device, comprising: Control center, data collection module, route planning module, cruise monitoring module, image processing and recognition module, ultrasonic data analysis and recognition module, comprehensive evaluation module and alarm module; characterized in that: Data collection module: collects the design drawings of the bridge steel structure, including the material, specifications and welding requirements of the steel structure, and collects images of welding defects; annotates the data as a reference sample; Route planning module: Based on the design drawings, GIS and CAD technology are used to build a three-dimensional model of the bridge steel structure. Considering the parameter factors of the drone, combined with the obstacle distribution in the three-dimensional model, the optimal flight path is calculated through algorithms; Cruise monitoring module: including drones, infrared thermal imagers, high-definition cameras and LED lights. According to the planned route, the drones carry high-definition cameras and infrared thermal imagers to conduct a full range of welding quality inspections on the bridges. Image processing and recognition module: pre-processes, extracts features and recognizes images collected by high-definition cameras to identify welding quality problems of bridge steel structures, including cracks, lack of fusion, pores, slag inclusions and deformation; and identifies abnormal situations where material dimensions do not meet requirements; Infrared data analysis and identification module: Process and analyze the infrared wave data collected by the infrared thermal imager to identify welding quality problems of bridge steel structures, including cracks, lack of fusion, pores, slag inclusions and deformation; Comprehensive evaluation module: Combines the analysis results of the image processing recognition module and the infrared data analysis recognition module to comprehensively evaluate the overall welding quality of the steel structure; evaluates the impact of welding quality on the overall safety of the bridge based on defect type, location, and severity factors; Alarm module: including an alarm, which will sound an alarm in time when welding quality problems are detected; Control center: network connected with data collection module, route planning module, cruise monitoring module, image processing and recognition module, ultrasonic data analysis and recognition module, comprehensive evaluation module and alarm module.

Images