Large steel box girder flaw detection method and related product

Abstract

The invention relates to the field of nondestructive testing, in particular to a flaw detection method for a large steel box girder and a related product, and the flaw detection method comprises the following steps: importing a three-dimensional model of the large steel box girder, and analyzing the three-dimensional model to obtain geometric feature information of the steel box girder; generating a scanning path of the ultrasonic probe based on the geometric feature information; an ultrasonic probe is controlled to scan the large steel box girder according to the scanning path, and ultrasonic echo signals are collected; the ultrasonic echo signals are processed, and the welding defect positions of the steel box girder are recognized and positioned; according to the method, the three-dimensional model of the steel box girder is imported and analyzed, automatic extraction of the detection area and the welding seam information is achieved, the optimized probe scanning path is generated by applying the self-adaptive path planning algorithm, efficient and full-coverage detection of the large complex structural part is achieved, and missing detection is avoided.

Description

Technical Field

[0001] The present invention relates to the field of nondestructive testing, and in particular to a large steel box girder flaw detection method and related products. Background Art

[0002] Large steel box girder structures are widely used in bridges, buildings, ships, and other fields. Ensuring the welding quality of large steel box girders is crucial to the safety of the structure. Currently, the inspection of the welding quality of large steel box girders mainly relies on non-destructive testing technology.

[0003] Nondestructive testing (NDT) technology can detect internal or surface defects without damaging the object being tested. Common NDT methods include radiographic testing, ultrasonic testing, magnetic particle testing, and penetrant testing. Ultrasonic testing, among others, has been widely used in weld defect detection due to its harmlessness to the human body, strong penetrant capability, high sensitivity, and ease of operation.

[0004] Existing ultrasonic testing technologies for large steel box girders typically rely on manual or semi-automated methods. Manual testing is labor-intensive, inefficient, and susceptible to test results influenced by operator experience. Furthermore, it is difficult to fully inspect large components. While semi-automated testing equipment improves testing efficiency to a certain extent, it often suffers from insufficient automation, poor adaptability to complex structures, and low intelligence levels, making it difficult to meet the demand for efficient, accurate, and automated testing of large steel box girders. Summary of the Invention

[0005] The technical problem to be solved by the present invention is that the existing large steel box girder welding quality inspection technology has the problems of high labor intensity and low efficiency in manual operation. The purpose is to provide a large steel box girder flaw detection method and related products, which realize the automated, intelligent, high-precision and high-efficiency detection of large steel box girder welding defects.

[0006] The present invention is achieved through the following technical solutions: A large steel box girder flaw detection method, comprising: Import the 3D model of a large steel box girder and analyze the 3D model to obtain the geometric characteristics of the steel box girder; generating a scanning path of the ultrasonic probe based on the geometric feature information; Control the ultrasonic probe to scan the large steel box girder along the scanning path and collect ultrasonic echo signals; The ultrasonic echo signal is processed to identify and locate the welding defects of the steel box girder.

[0007] Optionally, the method for obtaining geometric characteristic information of the steel box girder includes: Read the 3D model of large steel box girders. The 3D model file is in STEP or IGES format. Parse the entities and relationships in the 3D model file to obtain the geometric and topological information of the surface, the geometric and topological information of the edges, and the coordinate information of the vertices of the large steel box girder; Based on the geometric and topological information of the surface, the geometric and topological information of the edge, and the coordinate information of the vertex, the plate thickness of the large steel box girder is calculated, the weld location and weld type are identified, and the reinforcement position is identified.

[0008] Alternatively, the plate thickness is calculated as: Calculate the distance between two opposite parallel faces of a large steel box girder; The method for identifying the weld position is: Find adjacent faces of large steel box girders; Calculate the angle between adjacent faces; If the included angle is less than a first preset threshold and the adjacent faces are connected by a common edge, it is determined to be a potential weld; Determining the thickness of geometric bodies on both sides of the potential weld, and if the thickness of the geometric bodies is less than a second preset threshold, determining that the potential weld is a weld; The method for determining the weld type is as follows: set a1, a2, and a3 to the preset error angle values; If the angle between the surfaces on both sides of the weld is 180±a1 degrees, the weld is judged to be a butt weld; If the angle between the surfaces on both sides of the weld is 90±a2 degrees, the weld is judged to be a fillet weld; If the edge of the face on one side of the weld is connected to the inside of the other side, and the angle between the two faces is 90 ± a3 degrees, the weld is judged to be a T-weld; The method for identifying the position of the reinforcement is: Calculating the aspect ratio of each face in the three-dimensional model, and determining faces with an aspect ratio greater than a third preset threshold as thin plate structures; Find thin plate structures in large steel box girders; Determine whether the thin plate structure is connected to the main plate of the large steel box girder; If the thin plate structure is connected to the main plate, it is determined that the thin plate structure is a reinforcing rib.

[0009] Optionally, the method for generating a scanning path of an ultrasonic probe based on geometric feature information includes: Geometric feature information includes plate thickness, weld location, weld type, and rib location; Determine the weld area that requires ultrasonic testing based on the weld location, and select the appropriate ultrasonic probe and scanning method based on the weld type; Based on the weld area, plate thickness, ultrasonic probe, scanning method, and rib location, the environment is modeled and function defined to generate a scanning path for the ultrasonic probe that avoids ribs and other non-inspection areas.

[0010] Optionally, the method of environment modeling and function definition includes: Discretize the inspection area of ​​large steel box girder into a grid map ,in, is a grid node set, each grid node The coordinates are , is the set of edges connecting adjacent grid nodes; Determine the grid nodes Status attributes ,in, Indicates idle time. Indicates that it is occupied by obstacles. Indicates the weld area; Determine the grid nodes Normal vector of the surface ; Define the slave grid node Move to adjacent grid node The cost function ,in, Grid nodes To Grid Node The Euclidean distance of Adjust the cost for the probe pose, is the penalty coefficient determined by the weld type, To avoid the cost of obstacles, is the weld priority cost, is the weight coefficient; Define the slave grid node To the target node The heuristic function ,in, is the unit distance movement cost, The distance from the current node to the target node in three directions.

[0011] Optionally, the method for generating the scanning path includes: initialization: Create an OPEN list for storing nodes to be evaluated and a CLOSED list for storing evaluated nodes; The starting node Add to the OPEN list and set its initial price , total cost estimate , is the starting node To the target node The heuristic function of is the initial cost function value, Passing the starting node The total cost estimate of the path; Iterative search: Loop through the following steps until the OPEN list is empty or the target node is found: Select the total cost estimate from the OPEN list The node with the smallest value , and judge Is it the target node? If yes, the search is completed; if not, Move from the OPEN list to the CLOSED list and continue searching; Traversal All adjacent nodes , and do the following: if In the CLOSED list, it is skipped; if If it is not in the CLOSED list, the calculation starts from the starting point arrive The actual cost , From the grid node Move to adjacent grid node The cost function of is the starting node To Node The actual cost function of if If it is not in the OPEN list, Add to OPEN list, set , , and record The parent node is ; if In the OPEN list, and , then update , , and update The parent node is ; Path backtracking: From the target node Start by tracing back to the starting node based on the parent node recorded by each node , generate preliminary paths; Path Smoothing: The generated preliminary path is smoothed using Bezier curves to obtain the scanning path.

[0012] Optionally, the method of performing an ultrasound scan includes: The scanning device includes: a mobile platform, a mechanical arm and an ultrasonic probe, wherein the mechanical arm is mounted on the mobile platform; When performing ultrasonic scanning, the mobile platform and the robotic arm are controlled to move so that the ultrasonic probe moves along the scanning path; According to the type of weld, select the corresponding ultrasonic probe and scanning method: If the weld type is a butt weld, select an oblique probe to transmit and receive ultrasonic waves and perform a transverse scan; If the weld type is a fillet weld, select a dual crystal probe or a TOFD probe to transmit and receive ultrasonic waves and perform a longitudinal scan; If the weld type is a T-type weld, a climbing probe or a phased array probe is selected to transmit and receive ultrasonic waves and perform a transverse or longitudinal scan.

[0013] Optionally, the method for identifying the location of a welding defect includes: Perform discrete wavelet transform on ultrasonic echo signal ,in, is the wavelet coefficient, is the scale parameter, is the translation parameter, is the complex conjugate of the wavelet basis function, is the signal length, is the ultrasonic echo signal, For time; Determine the scale The wavelet coefficients Threshold ,in, For scale The standard deviation of the noise, For scale The number of wavelet coefficients under ; Use soft threshold function to process wavelet coefficients ; The processed wavelet coefficients Perform inverse discrete wavelet transform to obtain the denoised signal ; Construct TCG curve and perform noise reduction on the signal Compensate and obtain the final echo signal ; The final echo signal is extracted by short-time Fourier transform and continuous wavelet transform The time-frequency characteristics in ; Input the time-frequency features into the trained deep residual time-frequency network and output the feature vector ; The Softmax function is used to transform the feature vector Classify and calculate The probability value of a list ,in, is the actual number of defect categories, , Indicates no defects, Indicates various defect categories, For category The transposed matrix of the weight vector, For category The bias, For category The transposed matrix of the weight vector, For category Bias; Setting the confidence threshold ,like Then discard the classification results, if , then its corresponding As defect type; After completing the defect type identification on the scanning path, the defect location is performed on the weld with defects.

[0014] Optionally, the method for locating the welding defect position includes: The final echo signal Perform continuous wavelet transform to obtain wavelet coefficients , and calculate the wavelet energy distribution ,in, is the scale parameter, is the translation parameter; Select the time point corresponding to the maximum energy value as the arrival time of the defect echo ,in, For all scale parameters The maximum value of the wavelet energy distribution, is the translation parameter corresponding to the maximum value ; Calculate the depth of welding defect location ,in, is the ultrasonic wave propagation velocity in the large steel box girder material; Obtain the coordinates of the ultrasonic probe in the large steel box girder coordinate system and attitude angle ,in, is the roll angle, is the pitch angle, is the yaw angle; Calculate the coordinates of the defect position in the probe coordinate system , ; Convert the defect position coordinates from the probe coordinate system to the large steel box girder coordinate system to obtain the welding defect position coordinates , ,in, is the rotation matrix determined by the attitude angle, ,in, For around The rotation matrix of the axis, For around The rotation matrix of the axis, For around The rotation matrix about the axis.

[0015] A computer program product comprises a computer program / instruction. When the computer program / instruction is executed by a processor, the method described above is implemented.

[0016] Compared with the prior art, the present invention has the following advantages and beneficial effects: By importing and analyzing a three-dimensional model of a steel box girder, this method automatically extracts information about the inspection area and welds. By using an adaptive path planning algorithm to generate an optimized probe scanning path, this method achieves efficient, full-coverage inspection of large, complex structural components, preventing missed inspections. By constructing a deep residual time-frequency network and combining multiple time-frequency domain feature analysis methods to process ultrasonic echo signals, this method enables comprehensive detection and assessment of different defect types. Ultimately, this reduces labor costs and improves inspection efficiency and consistency. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] The accompanying drawings illustrate exemplary embodiments of the present invention and, together with the description thereof, are used to explain the principles of the present invention. These drawings are included to provide a further understanding of the present invention, and the accompanying drawings are included in and constitute a part of this specification and do not constitute a limitation of the embodiments of the present invention.

[0018] Figure 1 It is a flow chart of a large steel box girder flaw detection method according to the present invention.

[0019] Figure 2 4 is a flow chart of the scanning path generation method according to the present invention. DETAILED DESCRIPTION

[0020] To make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the relevant content and are not intended to limit the present invention.

[0021] It should also be noted that, for the convenience of description, only the parts related to the present invention are shown in the drawings.

[0022] In the absence of conflict, the embodiments and features of the embodiments of the present invention can be combined with each other. The present invention will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0023] Example 1, as Figure 1 As shown, a large steel box girder flaw detection method is provided, comprising: Model import and analysis: Import the 3D model of a large steel box girder and analyze the 3D model to obtain the geometric feature information of the steel box girder.

[0024] 3D models are typically generated in formats such as STEP and IGES using CAD (Computer-Aided Design) software. Once imported, the model is automatically parsed to extract the geometric characteristics of the steel box girder, including the shape, size, position, and orientation of each component (e.g., plates, welds, and reinforcements).

[0025] Scanning path generation: Generate the scanning path of the ultrasonic probe based on geometric feature information.

[0026] The path planning process considers several factors, such as the location and type of welds to be inspected; the overall structure of the steel box girder to avoid collisions between the probe and the girder; the probe's range of motion and position adjustment capabilities; and the inspection efficiency and coverage requirements.

[0027] Scanning and signal acquisition: Control the ultrasonic probe to scan the large steel box girder along the scanning path and collect ultrasonic echo signals.

[0028] The scanning device (e.g., a mobile robot with a robotic arm) is controlled to move the ultrasonic probe along the generated scanning path. As the probe moves, it transmits ultrasonic pulses into the interior of the steel box girder and receives ultrasonic echo signals reflected from the interior. These echo signals contain information about the internal structure of the steel box girder, including any possible weld defects.

[0029] Signal processing and defect identification and positioning: Process the ultrasonic echo signal to identify and locate the welding defect position of the steel box girder.

[0030] This typically involves signal preprocessing (such as noise reduction, filtering, and gain compensation), feature extraction, defect identification, and defect location. By analyzing the characteristics of the echo signal, it is possible to determine the presence of a defect, its type (such as pores, cracks, and lack of fusion), and its location and size.

[0031] This embodiment works by exploiting the reflection characteristics of ultrasonic waves propagating within a material to detect defects. By importing and analyzing a computer-aided design (CAD) model, geometric information about the steel box girder is acquired, enabling automated preparation for inspection. An intelligent path planning algorithm generates an optimized probe scanning path. Automated scanning and signal processing techniques enable the detection of weld defects in large steel box girders.

[0032] Example 2: This example details how to extract key geometric features for flaw detection from a 3D CAD model of a large steel box girder, including plate thickness, weld location and type, and rib position. The entire process is divided into three main stages: model reading, model parsing, and feature information calculation.

[0033] Methods for obtaining geometric characteristic information of steel box girder include: Read the 3D model of large steel box girders. The 3D model file is in STEP or IGES format. Parse the entities and relationships in the 3D model file to obtain the geometric and topological information of the surface, the geometric and topological information of the edges, and the coordinate information of the vertices of the large steel box girder.

[0034] A face's geometric information describes its shape (e.g., plane, curved surface), size, and position. Its topological information describes its connections with other faces, edges, and vertices. Edge geometric information describes its shape (e.g., straight line, curve), length, and position. Its topological information describes its connections with other faces and vertices. Vertex coordinate information describes its position in three-dimensional space.

[0035] Based on the geometric and topological information of faces, edges, and vertex coordinates, the thickness of large steel box girders is calculated, weld locations and types are identified, and reinforcement locations are identified. Specific calculation methods include: The plate thickness is calculated by calculating the distance between two relatively parallel surfaces of a large steel box girder; The weld location identification method is as follows: find adjacent faces of a large steel box girder; calculate the angle between the adjacent faces; if the angle is less than a first preset threshold and the adjacent faces are connected by a common edge, it is determined to be a potential weld; determine the thickness of the geometric bodies on both sides of the potential weld; if the thickness of the geometric bodies is less than a second preset threshold, the potential weld is determined to be a weld; The method for determining the weld type is: set a1, a2, and a3 to preset error angle values; generally set within 10 degrees.

[0036] If the angle between the surfaces on both sides of the weld is 180±a1 degrees, the weld is judged to be a butt weld; If the angle between the surfaces on both sides of the weld is 90±a2 degrees, the weld is judged to be a fillet weld; If the edge of the face on one side of the weld is connected to the inside of the other side, and the angle between the two faces is 90 ± a3 degrees, the weld is judged to be a T-weld; The method for identifying the position of the reinforcement is: Calculating the aspect ratio of each face in the three-dimensional model, and determining faces with an aspect ratio greater than a third preset threshold as thin plate structures; Find thin plate structures in large steel box girders; Determine whether the thin plate structure is connected to the main plate of the large steel box girder; If the thin plate structure is connected to the main plate, the thin plate structure is determined to be a reinforcement rib.

[0037] The working principle of this embodiment is to use the data structure and geometric modeling kernel of the CAD model to access and extract the geometric and topological information of the model through a programming interface, and automatically extract key information such as plate thickness, weld position, weld type and rib position from the CAD model through a series of geometric calculations and logical judgments.

[0038] Example 3 illustrates how to generate an ultrasonic probe's scanning path based on the previously extracted geometric features of a steel box girder (plate thickness, weld location, weld type, and rib location). This process consists of two main parts: path planning preparation (environment modeling and function definition) and path generation.

[0039] like Figure 2 As shown, the method for generating a scanning path of an ultrasonic probe based on geometric feature information includes: Geometric feature information includes plate thickness, weld location, weld type, and rib location; Determine the weld area that requires ultrasonic testing based on the weld location, and select the appropriate ultrasonic probe and scanning method based on the weld type; Based on the weld area, plate thickness, ultrasonic probe, scanning method, and rib location, the environment is modeled and function defined to generate a scanning path for the ultrasonic probe that avoids ribs and other non-inspection areas.

[0040] The preliminary preparations for path planning, namely the methods of environment modeling and function definition, include: Divide the inspection area of ​​the large steel box girder (usually the weld and its surrounding area) into small, regular grids, that is, discretize the inspection area of ​​the large steel box girder into a grid map ,in, is a grid node set, each grid node The coordinates are , is the set of edges connecting adjacent grid nodes.

[0041] Determine the grid nodes Status attributes ,in, Indicates idle, the probe can pass; Indicates that the probe cannot pass through due to an obstacle, such as a reinforcement rib; Indicates the weld area, which requires intensive inspection.

[0042] Determine the grid nodes Normal vector of the surface ; that is, the orientation of the surface where the grid is located. For the weld area, the normal vector represents the orientation of the weld surface.

[0043] Define the slave grid node Move to adjacent grid node The cost function ,in, Grid nodes To Grid Node The Euclidean distance of Adjust the cost for the probe pose, is the penalty coefficient determined by the weld type, To avoid the cost of obstacles, is the weld priority cost, is the weight coefficient; Because welds may have different orientations, the probe needs to adjust its posture to ensure vertical ultrasonic incidence. Therefore, the probe posture adjustment cost is related to the angle between the two grid normal vectors and is affected by different weld types. It is adjusted through a penalty coefficient, which varies for different types of welds.

[0044] If the target grid is an obstacle, the obstacle avoidance cost is infinite. If the target grid is a weld area, the weld priority cost is negative (encouraging the probe to scan the weld first).

[0045] Define the slave grid node To the target node The heuristic function ,in, is the unit distance movement cost, The distance from the current node to the target node in three directions.

[0046] After completing the environment modeling and function definition (cost function and heuristic function), a search algorithm is used to generate the scanning path. The generation method includes: initialization: Create an OPEN list for storing nodes to be evaluated and a CLOSED list for storing evaluated nodes; The starting node Add to the OPEN list and set its initial price , total cost estimate ; The cost from the starting node to itself is 0, and the estimated value is the estimated cost from the starting node to the target node. is the starting node To the target node The heuristic function of is the initial cost function value, Passing the starting node The total cost estimate of the path.

[0047] Iterative search: Loop through the following steps until the OPEN list is empty (meaning no path is found) or the target node is found (meaning a path is found): Select the total cost estimate from the OPEN list The node with the smallest value , and judge Is it the target node? If yes, the search is completed, the iterative search ends, and the path backtracking phase begins. Move from the OPEN list to the CLOSED list (indicating that this node has been evaluated and no longer needs to be considered), and continue searching; Traversal All adjacent nodes , and do the following: if In the CLOSED list, it is skipped (because it has already been evaluated); if If it is not in the CLOSED list, the calculation starts from the starting point arrive The actual cost , From the grid node Move to adjacent grid node The cost function of is the starting node To Node The actual cost function.

[0048] Next, there are two situations: if If it is not in the OPEN list, Add to OPEN list, set , , and record The parent node is ; if In the OPEN list, and , then update , , and update The parent node is ; Path backtracking: From the target node Start by tracing back to the starting node based on the parent node recorded by each node , generate preliminary paths; Path Smoothing: Due to the discrete nature of the raster map, the initial path obtained by backtracking is usually composed of a series of broken line segments and is not smooth enough. Therefore, Bezier curves are used to smooth the generated initial path to obtain the scanning path.

[0049] The method in this embodiment works by using an iterative search algorithm based on a grid map, a cost function, and a heuristic function to find the optimal path from the starting node to the target node. After finding the preliminary path, it is smoothed using a Bezier curve to obtain the final scan path.

[0050] Embodiment 4: This embodiment describes the use of a scanning device to perform actual ultrasonic scanning operations according to the previously generated scanning path and the determined weld type.

[0051] Ultrasound scans are performed by: The scanning device consists of a mobile platform, a robotic arm, and an ultrasonic probe. The robotic arm is mounted on the mobile platform. The mobile platform provides overall mobility, enabling the robotic arm to reach different locations on the steel box girder. An omnidirectional mobile robot chassis (e.g., Mecanum wheel chassis, differential wheel chassis, etc.) should be used and equipped with a positioning system (e.g., LiDAR, vision navigation, etc.) to achieve precise autonomous navigation and positioning.

[0052] The robotic arm is installed on a mobile platform, providing precise positioning and posture adjustment capabilities, enabling the probe to move along the scanning path. A six-axis or seven-axis industrial robotic arm is used, and the joint movement is precisely controlled by the control system to achieve precise movement and posture adjustment of the probe on the surface of the steel box girder.

[0053] Different types of probes (angle probe, dual crystal probe, TOFD probe, climbing probe, phased array probe) are selected according to different weld types (butt joint, fillet joint, T-type), and the ultrasonic probe is installed at the end of the robotic arm to transmit and receive ultrasonic waves.

[0054] During ultrasonic scanning, the mobile platform and robotic arm are controlled to move the ultrasonic probe along the scanning path. The mobile platform is responsible for large-scale movement, transporting the robotic arm to different inspection areas of the steel box girder. The robotic arm is responsible for fine movements, accurately moving the probe along the scanning path and adjusting the probe's posture to ensure that the ultrasonic waves enter the weld at the optimal angle.

[0055] According to the type of weld, select the corresponding ultrasonic probe and scanning method: If the weld type is a butt weld, select an angle probe to transmit and receive ultrasonic waves and perform a transverse scan. The ultrasonic waves generated by the angle probe are incident on the steel box girder at a certain angle, which can effectively detect defects in the butt weld. The transverse scanning probe moves perpendicular to the direction of the weld.

[0056] For fillet welds, a dual element probe or TOFD probe is used to transmit and receive ultrasonic waves, performing longitudinal scanning. A dual element probe uses one element to transmit ultrasonic waves and the other to receive them. This makes it suitable for detecting flaws near the root and surface of fillet welds. TOFD probes utilize diffracted waves from the end points of the flaw for detection. Longitudinal scanning involves moving the probe parallel to the weld's direction.

[0057] If the weld is a T-weld, a ramp probe or phased array probe is selected to transmit and receive ultrasonic waves, performing transverse or longitudinal scanning. A ramp probe propagates ultrasonic waves along the groove of the T-weld. A phased array probe electronically controls the deflection and focus of the ultrasonic beam, and the scanning method depends on the geometry of the T-weld and the possible location of the defect.

[0058] The working principle is to control the scanning device (mobile platform and robotic arm) to move the ultrasonic probe according to the pre-planned scanning path, and select the appropriate probe and scanning method according to the weld type during the movement.

[0059] Embodiment 5: This embodiment provides a method for processing and analyzing collected ultrasonic echo signals to identify whether there are defects and the type of defects.

[0060] Methods for identifying welding defect locations include: Perform discrete wavelet transform on ultrasonic echo signal ,in, is the wavelet coefficient, is the scale parameter, is the translation parameter, is the complex conjugate of the wavelet basis function, is the signal length, is the ultrasonic echo signal, For time; Determine the scale The wavelet coefficients Threshold ,in, For scale The standard deviation of the noise, For scale The number of wavelet coefficients under ; Use soft threshold function to process wavelet coefficients ; The processed wavelet coefficients Perform inverse discrete wavelet transform to obtain the denoised signal ; Since ultrasonic waves will attenuate during propagation, the echo signal of deep defects will be weak. In order to compensate for this attenuation, it is necessary to perform time gain compensation on the noise-reduced signal. Construct a TCG curve and calculate the noise-reduced signal. Compensate and obtain the final echo signal .

[0061] The final echo signal is extracted by short-time Fourier transform and continuous wavelet transform The time-frequency characteristics in the data; such as time-frequency energy distribution, instantaneous frequency, instantaneous bandwidth, wavelet entropy, time-frequency ridges, etc.

[0062] Input the time-frequency features into the trained deep residual time-frequency network and output the feature vector ; The deep residual time-frequency network is a deep learning model that is good at processing time series data and image data. It consists of multiple residual blocks, each of which contains a convolutional layer, a batch normalization layer, a ReLU activation function, and an attention mechanism.

[0063] The Softmax function is used to transform the feature vector For classification, the Softmax function can convert the feature vector into a probability distribution, which represents the probability that the signal belongs to each defect category (including the "no defect" category).

[0064] Calculate the The probability value of a list ,in, is the actual number of defect categories, , Indicates no defects, Indicates various defect categories, For category The transposed matrix of the weight vector, For category The bias, For category The transposed matrix of the weight vector, For category Bias; Setting the confidence threshold ,like Then discard the classification results, if , then its corresponding As defect type; if If the value is equal to 0, it is judged as no defect, and if it is equal to other values, it means there is a corresponding defect.

[0065] After the defect type identification on the scanning path is completed, the defect location of the weld with defects is performed. The location location methods of the welding defects include: The final echo signal Perform continuous wavelet transform to obtain wavelet coefficients , and calculate the wavelet energy distribution ,in, is the scale parameter, is the translation parameter; Select the time point corresponding to the maximum energy value as the arrival time of the defect echo ,in, For all scale parameters The maximum value of the wavelet energy distribution, is the translation parameter corresponding to the maximum value ; Calculate the depth of welding defect location ,in, is the ultrasonic wave propagation velocity in the large steel box girder material; Obtain the coordinates of the ultrasonic probe in the large steel box girder coordinate system and attitude angle ,in, is the roll angle, is the pitch angle, is the yaw angle; the steel box girder coordinate system takes a corner point of the large steel box girder as the origin, and the length, width and height directions of the large steel box girder as the coordinate axes.

[0066] Calculate the coordinates of the defect position in the probe coordinate system , The probe coordinate system is established with the center of the ultrasonic probe as the origin. The z-axis of the probe coordinate system is perpendicular to the ultrasonic probe, and the plane where the x-axis and y-axis of the probe coordinate system are located is parallel to the ultrasonic probe.

[0067] Convert the defect position coordinates from the probe coordinate system to the large steel box girder coordinate system to obtain the welding defect position coordinates , ,in, is the rotation matrix determined by the attitude angle, ,in, For around The rotation matrix of the axis, For around The rotation matrix of the axis, For around The rotation matrix about the axis.

[0068] Finally, a method for constructing a TCG curve is provided.

[0069] Construct the initial TCG curve, ,in, is the propagation time of the ultrasonic wave, is the speed of ultrasound in the material, is the frequency of the ultrasound, A constant for adjusting the overall gain level, is the frequency-dependent material attenuation coefficient.

[0070] Make a standard test block containing a flat-bottom hole of known depth (or use an actual workpiece with a flaw of known depth).

[0071] Use an ultrasonic probe to scan the test block and record the echo signals of flat-bottom holes of different depths.

[0072] Measure the amplitude of each flat bottom hole echo signal , It is The echo time corresponding to a flat-bottom hole.

[0073] Calculated according to the theoretical model, without considering the theoretical amplitude of attenuation .

[0074] Calculate the amplitude difference .

[0075] Perform curve fitting (e.g., using polynomial fitting) to obtain a calibration curve .

[0076] The calibration curve is superimposed on the initial TCG curve to obtain the final TCG curve: .

[0077] Perform gain compensation .

[0078] Embodiment 6. A computer-readable storage medium stores a computer program. When the computer program is executed by a processor, the steps of the above-mentioned antenna interface unit testing method are implemented.

[0079] Without loss of generality, computer-readable media may include computer storage media and communication media. Computer storage media includes volatile and non-volatile, removable and non-removable media implemented in any method or technology for storing information such as computer-readable instruction data structures, program modules, or other data. Computer storage media includes RAM, ROM, EPROM, EEPROM, flash memory or other solid-state storage technologies, CD-ROM, DVD or other optical storage, magnetic cassettes, magnetic tape, magnetic disk storage or other magnetic storage devices. Of course, those skilled in the art will appreciate that computer storage media is not limited to the aforementioned types. The aforementioned system memory and mass storage devices may be collectively referred to as memory.

[0080] A computer program product comprises a computer program / instruction, which implements the steps of any one of the above methods when executed by a processor.

[0081] A computer program product consists of a computer program or set of instructions for performing specific tasks or implementing specific functions. These programs or instructions are designed to be executed by a processor, thereby completing a series of predefined steps or operations. The program product may be stored in various forms of computer storage media, such as memory, hard disks, solid-state drives, optical disks, or other digital storage devices. It may exist as compiled binary code or as a script or bytecode executable by an interpreter. Through carefully designed algorithms and logical instructions, the program product enables the processor to process data in a specific order and manner, completing various functions such as data analysis, user interaction, and device control.

[0082] In the description of this specification, the description with reference to the terms "one embodiment / method", "some embodiments / methods", "example", "specific example", or "some examples" means that the specific features, structures, materials or characteristics described in conjunction with the embodiment / method or example are included in at least one embodiment / method or example of the present application. In this specification, the schematic expressions of the above terms do not necessarily refer to the same embodiment / method or example. Moreover, the specific features, structures, materials or characteristics described may be combined in an appropriate manner in any one or more embodiments / methods or examples. In addition, those skilled in the art may combine and combine different embodiments / methods or examples described in this specification and the features of different embodiments / methods or examples, unless they are contradictory.

[0083] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of the technical features being referred to. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of such features. Throughout the description of this application, "plurality" means at least two, for example, two, three, etc., unless otherwise specifically defined.

[0084] It should be understood by those skilled in the art that the above embodiments are merely for the purpose of illustrating the present invention clearly, and are not intended to limit the scope of the present invention. For those skilled in the art, other changes or modifications may be made based on the above invention, and these changes or modifications are still within the scope of the present invention.

Claims

1. A large steel box girder flaw detection method, characterized in that: include: Import the 3D model of a large steel box girder and analyze the 3D model to obtain the geometric characteristics of the steel box girder; generating a scanning path of the ultrasonic probe based on the geometric feature information; Control the ultrasonic probe to scan the large steel box girder along the scanning path and collect ultrasonic echo signals; The ultrasonic echo signal is processed to identify and locate the welding defects of the steel box girder.

2. A large steel box girder flaw detection method according to claim 1, characterized in that: Methods for obtaining geometric characteristic information of steel box girder include: Read the 3D model of large steel box girders. The 3D model file is in STEP or IGES format. Parse the entities and relationships in the 3D model file to obtain the geometric and topological information of the surface, the geometric and topological information of the edges, and the coordinate information of the vertices of the large steel box girder; Based on the geometric and topological information of the surface, the geometric and topological information of the edge, and the coordinate information of the vertex, the plate thickness of the large steel box girder is calculated, the weld location and weld type are identified, and the reinforcement position is identified.

3. A large steel box girder flaw detection method according to claim 2, characterized in that: The calculation method of plate thickness is: Calculate the distance between two opposite parallel faces of a large steel box girder; The method for identifying the weld position is: Find adjacent faces of large steel box girders; Calculate the angle between adjacent faces; If the included angle is less than a first preset threshold and the adjacent faces are connected by a common edge, it is determined to be a potential weld; Determining the thickness of geometric bodies on both sides of the potential weld, and if the thickness of the geometric bodies is less than a second preset threshold, determining that the potential weld is a weld; The method for determining the weld type is as follows: set a1, a2, and a3 to the preset error angle values; If the angle between the surfaces on both sides of the weld is 180±a1 degrees, the weld is judged to be a butt weld; If the angle between the surfaces on both sides of the weld is 90±a2 degrees, the weld is judged to be a fillet weld; If the edge of the face on one side of the weld is connected to the inside of the other side, and the angle between the two faces is 90 ± a3 degrees, the weld is judged to be a T-weld; The method for identifying the position of the reinforcement is: Calculating the aspect ratio of each face in the three-dimensional model, and determining faces with an aspect ratio greater than a third preset threshold as thin plate structures; Find thin plate structures in large steel box girders; Determine whether the thin plate structure is connected to the main plate of the large steel box girder; If the thin plate structure is connected to the main plate, it is determined that the thin plate structure is a reinforcing rib.

4. A large steel box girder flaw detection method according to claim 1, characterized in that: The method for generating a scanning path of an ultrasonic probe based on geometric feature information includes: Geometric feature information includes plate thickness, weld location, weld type, and rib location; Determine the weld area that requires ultrasonic testing based on the weld location, and select the appropriate ultrasonic probe and scanning method based on the weld type; Based on the weld area, plate thickness, ultrasonic probe, scanning method, and rib location, the environment is modeled and function defined to generate a scanning path for the ultrasonic probe that avoids ribs and other non-inspection areas.

5. A large steel box girder flaw detection method according to claim 4, characterized in that: Methods for environment modeling and function definition include: Discretize the inspection area of ​​large steel box girder into a grid map ,in, is a grid node set, each grid node The coordinates are , is the set of edges connecting adjacent grid nodes; Determine the grid nodes Status attributes ,in, Indicates idle time. Indicates that it is occupied by obstacles. Indicates the weld area; Determine the grid nodes Normal vector of the surface ; Define the slave grid node Move to adjacent grid node The cost function ,in, Grid nodes To Grid Node The Euclidean distance of Adjust the cost for the probe pose, is the penalty coefficient determined by the weld type, To avoid the cost of obstacles, is the weld priority cost, is the weight coefficient; Define the slave grid node To the target node The heuristic function ,in, is the unit distance movement cost, The distance from the current node to the target node in three directions.

6. A large steel box girder flaw detection method according to claim 5, characterized in that: The scan path generation method includes: initialization: Create an OPEN list for storing nodes to be evaluated and a CLOSED list for storing evaluated nodes; The starting node Add to the OPEN list and set its initial price , total cost estimate ; is the starting node To the target node The heuristic function of is the initial cost function value, Passing the starting node The total cost estimate of the path; Iterative search: Loop through the following steps until the OPEN list is empty or the target node is found: Select the total cost estimate from the OPEN list The node with the smallest value , and judge Is it the target node? If yes, the search is completed; if not, Move from the OPEN list to the CLOSED list and continue searching; Traversal All adjacent nodes , and do the following: if In the CLOSED list, it is skipped; if If it is not in the CLOSED list, the calculation starts from the starting point arrive The actual cost , From the grid node Move to adjacent grid node The cost function of is the starting node To Node The actual cost function of if If it is not in the OPEN list, Add to OPEN list, set , , and record The parent node is , For nodes To the target node The heuristic function of is the starting node To any node The actual cost function, Passing node The total cost estimate of the path; if In the OPEN list, and , then update , , and update The parent node is ; Path backtracking: From the target node Start by tracing back to the starting node based on the parent node recorded by each node , generate preliminary paths; Path Smoothing: The generated preliminary path is smoothed using Bezier curves to obtain the scanning path.

7. A large steel box girder flaw detection method according to claim 3, characterized in that: Ultrasound scans are performed by: The scanning device includes: a mobile platform, a mechanical arm and an ultrasonic probe, wherein the mechanical arm is mounted on the mobile platform; When performing ultrasonic scanning, the mobile platform and the robotic arm are controlled to move so that the ultrasonic probe moves along the scanning path; According to the type of weld, select the corresponding ultrasonic probe and scanning method: If the weld type is a butt weld, select an oblique probe to transmit and receive ultrasonic waves and perform a transverse scan; If the weld type is a fillet weld, select a dual crystal probe or a TOFD probe to transmit and receive ultrasonic waves and perform a longitudinal scan; If the weld type is a T-type weld, a climbing probe or a phased array probe is selected to transmit and receive ultrasonic waves and perform a transverse or longitudinal scan.

8. A large steel box girder flaw detection method according to claim 1, characterized in that: Methods for identifying welding defect locations include: Perform discrete wavelet transform on ultrasonic echo signal ,in, is the wavelet coefficient, is the scale parameter, is the translation parameter, is the complex conjugate of the wavelet basis function, is the signal length, is the ultrasonic echo signal, For time; Determine the scale The wavelet coefficients Threshold ,in, For scale The standard deviation of the noise, For scale The number of wavelet coefficients under ; Use soft threshold function to process wavelet coefficients ; The processed wavelet coefficients Perform inverse discrete wavelet transform to obtain the denoised signal ; Construct TCG curve and perform noise reduction on the signal Compensate and obtain the final echo signal ; The final echo signal is extracted by short-time Fourier transform and continuous wavelet transform The time-frequency characteristics in ; Input the time-frequency features into the trained deep residual time-frequency network and output the feature vector ; The Softmax function is used to transform the feature vector Classify and calculate The probability value of a list ,in, is the actual number of defect categories, , Indicates no defects, Indicates various defect categories, For category The transposed matrix of the weight vector, For category The bias, For category The transposed matrix of the weight vector, For category Bias; Setting the confidence threshold ,like Then discard the classification results, if , then its corresponding As defect type, is the maximum value of the probability value; After completing the defect type identification on the scanning path, the defect location is performed on the weld with defects.

9. A large steel box girder flaw detection method according to claim 8, characterized in that: Methods for locating welding defects include: The final echo signal Perform continuous wavelet transform to obtain wavelet coefficients , and calculate the wavelet energy distribution ,in, is the scale parameter, is the translation parameter; Select the time point corresponding to the maximum energy value as the arrival time of the defect echo ,in, For all scale parameters The maximum value of the wavelet energy distribution, is the translation parameter corresponding to the maximum value ; Calculate the depth of weld defect locations ,in, is the ultrasonic wave propagation velocity in the large steel box girder material; Obtain the coordinates of the ultrasonic probe in the large steel box girder coordinate system and attitude angle ,in, is the roll angle, is the pitch angle, is the yaw angle; Calculate the coordinates of the defect position in the probe coordinate system , ; Convert the defect position coordinates from the probe coordinate system to the large steel box girder coordinate system to obtain the welding defect position coordinates , ,in, is the rotation matrix determined by the attitude angle, ,in, For around The rotation matrix of the axis, For around The rotation matrix of the axis, For around The rotation matrix about the axis.

10. A computer program product comprising a computer program / instructions, characterized in that When the computer program / instructions are executed by a processor, the method according to any one of claims 1 to 9 is implemented.

Images