A bogie frame weld joint detection method, system and terminal device

Abstract

This invention discloses a method, system, and terminal device for inspecting weld seams on a bogie frame. The method includes: determining a frame model and a weld seam model using point cloud information of the bogie frame; determining the motion paths of each component of the weld seam inspection device based on the frame model and weld seam model; determining the real-time paint thickness information of the frame in the area corresponding to the probe, and dynamically determining the target defect detection frequency based on the real-time paint thickness information, while simultaneously determining the probe lift-off distance using a distance control algorithm to ensure that the probe maintains a constant distance from the frame surface; controlling the probe to detect the frame based on the target defect detection frequency and the probe lift-off distance, determining defect information, and determining weld defect information based on the defect information. This invention, based on real-time paint thickness information, adjusts the detection frequency in real time and optimizes the probe lift-off distance, thereby achieving high-precision defect identification without the need for paint removal.

Description

Technical Field

[0001] This invention relates to the field of weld inspection technology, and in particular to a method, system and terminal equipment for inspecting welds on a bogie frame. Background Technology

[0002] The statements in this section are merely background information related to the present invention and do not necessarily constitute prior art.

[0003] As a core load-bearing component of rail vehicles, the quality of the welds on the bogie frame directly affects train operation safety. Traditionally, weld inspection of the bogie frame relies primarily on magnetic particle testing. However, this method requires removing the paint from the frame surface to eliminate interference with the inspection signal, necessitating significant time and resources for surface grinding before inspection and recoating with protective paint afterward. This not only significantly extends the maintenance cycle but also increases material and labor costs.

[0004] In the existing technology, although there are eddy current testing devices for components using a single robotic arm, and eddy current testing technology theoretically has the potential for non-contact testing, conventional eddy current equipment is limited by insufficient sensitivity of the probe to the paint layer, as well as by the complex structure of the frame and pipeline interference, which makes it difficult to plan the path of complex welds and has low efficiency in multi-area collaborative testing. As a result, it is difficult to realize engineering applications in the inspection of bogie frames. Furthermore, the single robotic arm system has obvious limitations in dynamic path planning, multi-degree-of-freedom collaboration, and anti-interference capabilities, making it difficult to meet the requirements of high-precision automated testing. Summary of the Invention

[0005] To address the aforementioned issues, this invention proposes a method, system, and terminal device for inspecting weld seams on a bogie frame. Based on real-time information about the paint layer thickness, the detection frequency and lifting distance are adjusted in real time, thereby achieving high-precision defect identification without the need for paint removal.

[0006] According to a first aspect of the present invention, a method for inspecting weld seams on a bogie frame is provided, comprising: Obtain the point cloud information of the steering frame, and determine the frame model and weld model using the point cloud information; The motion paths of each component of the weld inspection device are determined based on the framework model and the weld model. The weld inspection device includes at least a probe and a robotic arm. The real-time paint thickness information of the structure in the area corresponding to the probe is determined, and the target defect detection frequency is dynamically determined based on the real-time paint thickness information. At the same time, the probe lift-off distance is determined through a distance control algorithm to ensure that the probe and the structure surface maintain a constant distance. The probe is controlled to detect the frame according to the target defect detection frequency and probe lift-off distance to determine defect information, and weld defect information is determined according to the defect information.

[0007] As a further solution, point cloud information of the steering architecture is obtained, and the architecture model and weld model are determined using the point cloud information, specifically: An architecture model is constructed based on the point cloud information, and the architecture model is matched with a standard architecture model in a preset database to determine the matching degree; the preset database stores standard models of various types of steering architectures and corresponding weld seam models; When the matching degree is greater than the preset matching value, the target architecture model is determined, and then the weld model corresponding to the target architecture model is determined.

[0008] As a further solution, the motion paths of each component of the weld inspection device are determined based on the framework model and the weld model, specifically as follows: Based on the framework model, the spatial distribution of interference objects on it is determined. Combining the weld model and the spatial distribution of interference objects, the probe scanning trajectory is determined using B-spline curves. The robotic arm's motion trajectory is determined by inverse kinematics calculation based on the probe scanning trajectory and time optimization.

[0009] The gantry movement trajectory and the ground track movement trajectory are determined by the movement trajectory of the robotic arm.

[0010] As a further solution, the time optimization algorithm is to minimize the total motion time for the robotic arm to complete the probe scanning trajectory by adjusting the motion timing and velocity curve of each joint of the robotic arm, under the constraint that the angular acceleration and angular velocity of each joint of the robotic arm do not exceed the corresponding limit parameters of the robotic arm.

[0011] As a further step, after determining the robotic arm's motion trajectory, the following steps are also included: Construct an envelope sphere model for each link of the robotic arm; The robotic arm's motion trajectory is discretized by time step, and the real-time three-dimensional coordinates of each envelope sphere at each time step are calculated. All sphere pairs with potential collisions are identified by calculating the distance between the spheres, and all spheres that do not meet the detection distance parameters are identified by calculating the distance between the spheres and the structure surface. The final collision detection result is then determined. Based on the detection results, the movement trajectory of the robotic arm is adjusted according to a preset graded avoidance strategy.

[0012] As a further step, after determining the robotic arm's motion trajectory, the following steps are also included: Using the probe scanning trajectory, robotic arm movement trajectory, gantry movement trajectory, ground rail movement trajectory, and weld seam scanning parameters as constraints, the motion path of each component is determined by optimizing and adjusting the parameters through a preset objective function. The preset objective function is: ; in, , , These are the weighting coefficients, Total time The total energy consumption of all components. This represents the sum of joint accelerations.

[0013] As a further solution, the target defect detection frequency is dynamically determined based on the real-time paint layer thickness information, specifically as follows: A pre-calibrated table showing the correspondence between paint layer thickness and detection frequency is used to determine the corresponding detection frequency based on the real-time paint layer thickness.

[0014] As a further solution, determining the weld defect information based on the aforementioned defect information specifically involves: The weld defect information includes at least the crack depth; The defect information is processed using wavelet packet transform algorithm and adaptive noise cancellation algorithm to determine the target signal; The impedance plane coordinate set is determined based on the target information number, and the area of ​​the target ellipse is determined based on the impedance plane coordinate set using the impedance plane ellipse fitting formula. The crack depth is determined using the crack depth formula based on the area of ​​the target ellipse.

[0015] According to a second aspect of the present invention, a bogie frame weld inspection system is provided, comprising: The model determination module is configured to acquire point cloud information of the steering frame and determine the frame model and weld model through the point cloud information. The trajectory determination module is configured to determine the motion paths of each component of the weld detection device based on the framework model and the weld model, wherein the weld detection device includes at least a probe and a robotic arm; The detection parameter dynamic adjustment module is configured to determine the real-time paint layer thickness information of the structure corresponding to the probe, and dynamically determine the target defect detection frequency based on the real-time paint layer thickness information. At the same time, the probe lift-off distance is determined through a distance control algorithm to ensure that the probe and the structure surface maintain a constant distance. The defect information determination module is configured to control the probe to detect the frame according to the target defect detection frequency and the probe lift-off distance, determine the defect information, and determine the weld defect information according to the defect information.

[0016] According to a third aspect of the present invention, a terminal device is provided, comprising a processor and a memory, wherein the processor is configured to implement instructions; and the memory is configured to store a plurality of instructions adapted to be loaded by the processor and executed by the processor to perform the aforementioned method for detecting weld seams on a bogie frame.

[0017] Compared with the prior art, the beneficial effects of the present invention are: (1) When the eddy current probe of the robotic arm of the present invention performs defect scanning, it adjusts the detection frequency in real time based on the real-time information of the paint layer thickness and optimizes the probe lift-off distance, thereby achieving high-precision defect identification without the need for paint removal. This avoids the situation where traditional eddy current detection is easily affected by the fluctuation of paint layer thickness due to the probe lift-off effect, resulting in signal distortion.

[0018] (2) The present invention accurately determines the frame model and weld model through frame point cloud information, and determines the motion path of each component according to the frame model and weld model, combined with time optimization algorithm, and adjusts the motion trajectory of the robotic arm through collision detection algorithm, thereby minimizing the total motion time of the robotic arm under the premise of ensuring the stable and reliable operation of each component, ensuring the synergistic effect between multiple robotic arms, gantry and ground rail, and improving the weld detection efficiency of the frame.

[0019] Other features and advantages of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0020] Figure 1 This is a flowchart of the method for inspecting weld seams on a bogie frame in an embodiment of the present invention; Figure 2 This is a schematic diagram of the bogie frame weld inspection system in an embodiment of the present invention; Figure 3 This is a schematic diagram of the terminal device structure in an embodiment of the present invention. Detailed Implementation

[0021] It should be noted that the following detailed description is illustrative and intended to provide further explanation of the invention. Unless otherwise specified, all technical and scientific terms used in this invention have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0022] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the scope of exemplary embodiments according to the invention. As used herein, the singular form is intended to include the plural form as well, unless the context clearly indicates otherwise. Furthermore, it should be understood that when the terms "comprising" and / or "including" are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof.

[0023] Example 1 In one or more embodiments, a method for inspecting weld seams on a bogie frame is disclosed, combined with Figure 1 Specifically, it includes the following process: S101: Obtain the point cloud information of the steering frame, and determine the frame model and weld model through the point cloud information.

[0024] In this embodiment, a database is first constructed, which pre-stores a large number of CAD standard models of different types of bogie frames and their corresponding weld features. These standard models include: the three-dimensional geometry of the frame, the location, orientation, and dimensions of key welds, the spatial distribution of interference objects such as pipes / bolts, and optimization schemes for historical inspection parameters.

[0025] The scanning point cloud is aligned with the standard CAD model in the database using a point cloud registration algorithm. The matching degree is calculated. When the matching degree is greater than the preset matching value, it is determined to be the target structure model corresponding to the point cloud. The preset weld model of the target structure is automatically associated, including: the precise spatial coordinates of all welds to be detected, priority sorting (such as priority scanning of stressed welds), and parameters such as the corresponding robotic arm motion trajectory, so as to provide support for subsequent path planning.

[0026] If the matching fails, a review process will be triggered to determine the corresponding target architecture model. If the corresponding target architecture model does not exist in the database, a manual construction process will be triggered, and the constructed model and related parameters will be stored in the database as a new standard model.

[0027] S102: Determine the motion paths of each component of the weld inspection device based on the frame model and the weld model. The weld inspection device includes at least a probe and a robotic arm.

[0028] As an optional implementation, the weld inspection device also includes a gantry and a ground rail, with the gantry used to mount the robotic arm and the ground rail used to move the gantry.

[0029] This embodiment can adopt a single robotic arm working mode or a multi-robotic arm collaborative working mode. For example, one, two, or four robotic arms can be used to realize the synchronous scanning of the weld seams on the upper surface of the frame, with each robotic arm corresponding to a set of probes.

[0030] When multiple robotic arms work together, the area to be inspected on the frame is divided into zones according to the working range of the robotic arms. The scanning range of each robotic arm can be planned in advance to avoid interference. Then, the scanning path is confirmed by manually adjusting the limit movements of each robotic arm. At the same time, anti-collision limiters and laser real-time tracking limiters for multiple robotic arms are designed.

[0031] In this embodiment, the spatial distribution of interference objects (such as pipelines) on the framework model is determined. Combining the weld model and the spatial distribution of interference objects, the probe scanning trajectory is determined using B-spline curves. Then, the inverse solution operation is performed based on the probe scanning trajectory and time optimization calculation to determine the robot arm motion trajectory. The gantry motion trajectory and the ground rail motion trajectory are determined through the robot arm motion trajectory.

[0032] When multiple robotic arms work together, each robotic arm can determine the probe scanning trajectory using B-spline curves based on the weld seam model and the spatial distribution of interference objects within its corresponding scanning range, and calculate the motion trajectory of the corresponding robotic arm through inverse kinematics.

[0033] Specifically, B-spline curves naturally possess... Continuous or higher-order smoothness ensures smooth probe movement during operation, avoiding abrupt changes in speed and acceleration. This is crucial for guaranteeing the uniformity and repeatability of test data and protecting mechanical components. Furthermore, modifying a control point in the trajectory only affects a small segment of the curve, without altering the entire path. This allows for convenient fine-tuning of the trajectory without requiring a complete path redesign. The formula for a B-spline curve is as follows: ; in, As control points, For curve parameters, for j B-spline basis functions.

[0034] After determining the smooth scanning trajectory based on the B-spline curve, in order to make the center point of the probe reach this position on the trajectory and maintain the required posture, the angular velocity and angular acceleration of each joint of the robotic arm are determined by inverse kinematics, and then the motion trajectory of the robotic arm is determined.

[0035] As a specific implementation method, the process of determining the B-spline curve is as follows: S1021: Determine the basic constraints for curve generation. First, based on the previously acquired framework model and weld model, extract two types of key information as curve constraints: (1) Spatial distribution of interference objects: Identify the three-dimensional coordinates and dimensions of interference objects such as pipes and bolts from the framework model, and clarify the "prohibited areas" that the probe should avoid during scanning (such as the area within 50mm around the pipes). Weld inspection requirements: Obtain the precise spatial coordinates (start point, end point, inflection point), direction (straight line / curve), and inspection coverage width (e.g., extending 20mm on each side of the weld) of the target weld from the weld model, and clearly define the "necessary area" that the probe needs to cover.

[0036] (2) Constructing control points for the B-spline curve: Based on the above constraints, select several control points (denoted as ) within the "necessary region" and avoiding the "prohibited region". The selection of control points must meet the following requirements: ① Quantity and distribution requirements: One control point shall be set every 500 mm for straight sections of the weld, and one control point shall be set every 100 mm for curved sections (such as arc transitions) to ensure that the curve conforms to the direction of the weld. ② Avoidance requirements: The distance between all control points and the interfering object must be greater than or equal to the avoidance range (e.g., 50mm) to prevent the curve from getting too close to the interfering object; ③Endpoint constraints: The start and end points of the weld are forcibly set as control points to ensure that the scanning trajectory completely covers the weld.

[0037] S1022: Substitute the values ​​into the B-spline curve formula to generate the scanning trajectory. Use the B-spline curve formula to calculate the continuous path of the probe scanning trajectory. The formula is as follows: ; in, To scan any point (three-dimensional coordinates) on the trajectory, that is, the path points that the probe center point needs to follow; The curve parameter has a value range of [0,1], and iterates through it. It can generate a complete trajectory; This represents the j-th degree B-spline basis function, calculated using the basis function recursive formula; in the context of weld inspection, j The value is usually taken three times to ensure the trajectory. Continuous and smooth; Subtract 1 from the total number of control points.

[0038] S1022: Curve Verification and Fine-tuning: After generating the initial trajectory, it is necessary to verify whether it meets the following requirements: ① Weld coverage: Whether the trajectory completely covers the "necessary area" of the target weld, with no missed sections; ② Interference avoidance: Whether the distance between all points on the trajectory and the interfering object is greater than the avoidance range; ③ Smoothness: Whether the trajectory has no obvious inflection points (by calculating the curvature of adjacent path points, ensure that the rate of change of curvature is less than or equal to the requirements for the smoothness of the robotic arm's movement).

[0039] If the conditions are not met, adjust the positions of the control points in the corresponding area (e.g., move the control points near the interfering object), and resubmit them into the formula to generate the curve until all constraints are met.

[0040] As a specific implementation method, the inverse solution operation process is as follows: The probe scanning trajectory (i.e., the three-dimensional coordinate sequence of the probe center point) generated by the B-spline curve The inputs include time-step discretization (e.g., 10ms / step) and probe attitude requirements (e.g., probe perpendicular to the weld surface, attitude angle fixed at 0°). The motion parameters of each joint of the robotic arm, including the joint angular velocity at each time step, are used as inputs. With angular acceleration As output, the final motion trajectory of the robotic arm is formed.

[0041] Taking a six-degree-of-freedom robotic arm as an example, construct a kinematic model for the robotic arm: ; in, It is the pose matrix of the robotic arm end effector (i.e., the center point of the probe), which includes three-dimensional coordinates (corresponding to the scanning trajectory points) and attitude (corresponding to the preset probe attitude). These represent the rotation angles of the six joints of the robotic arm; The forward kinematics function is derived using the Denavit-Hartenberg parameter method and reflects the mapping relationship between joint rotation angle and end pose.

[0042] For each discrete point on the scanning trajectory (Corresponding to the kth time step), the end pose matrix is ​​known. (Depend on (Determined by the preset attitude), the solution satisfies Joint angle Specifically, an analytical method can be used to solve the problem: If the robotic arm has a standard structure (such as a ball-wrist type 6-DOF robotic arm), the analytical expression for the joint rotation angle can be derived through kinematic decomposition (decomposing the pose into inverse position and inverse attitude solutions), and then directly calculated. .

[0043] Since there may be multiple solutions for the inverse kinematics of the robotic arm (e.g., elbow pointing up / down), it is necessary to select the solutions that satisfy "joint rotation angles are within the limit range" (e.g., the range of each joint rotation angle is -180° to 180°) and "links have no self-interference" as valid solutions.

[0044] As a further implementation method, the process of determining the robotic arm's motion trajectory is further optimized using a time optimization algorithm to improve efficiency. The core of the time optimization algorithm is to minimize the total motion time for the robotic arm to complete the probe scanning trajectory while ensuring motion smoothness, under the premise of satisfying the robotic arm's hardware constraints.

[0045] The constraints of the time optimization algorithm are as follows: The algorithm must adhere to the physical limits of the robotic arm, namely the maximum acceleration of the joints ( ) and maximum speed ( The formula is expressed as follows: ; in, Indicates from the initial time Until the end time The integral of the integral is the total time of the robotic arm's motion. Indicates the first The absolute value of the angular acceleration of each joint. This indicates the maximum acceleration limit of the joint. Indicates the first The absolute value of the angular velocity of each joint. This indicates the maximum speed limit of the joint. Let be the acceleration of the i-th joint of the robotic arm. To ensure the joint speed is within the hardware's capabilities, avoid exceeding the limits that could lead to malfunctions or decreased accuracy.

[0046] Under the above constraints, by adjusting the motion timing of each joint of the robotic arm (such as the change in rotation angle between adjacent time steps) and the speed curve (such as using a trapezoidal speed curve), the total motion time of the robotic arm from the starting position to the ending position through the scanning trajectory is minimized, ensuring the collaborative efficiency of multiple robotic arms, gantry, and ground rail (the total detection time is determined by the longest motion time of each component).

[0047] By optimizing the calculation, the joint angular velocity at each time step is obtained. With angular acceleration The sequence of parameters for all time steps constitutes the motion trajectory of the robotic arm.

[0048] Finally, verify whether the robotic arm's motion trajectory meets the following requirements: ① End-effector pose accuracy: When the robotic arm moves along the trajectory, the actual pose of the end-effector (probe center point) and the scanning trajectory point The deviation is less than or equal to the required detection accuracy (e.g., less than 0.1 mm). ② Joint parameter compliance: all time steps , None of them exceeded the hardware limits; ③ No collision risk: Collision detection is performed in conjunction with the subsequent "envelope sphere model". If a collision exists, the inverse solution step is returned, and the joint angle solution is adjusted or the parameters are optimized.

[0049] As a further implementation, after obtaining the motion trajectory of each robotic arm, an envelope sphere model is constructed for each link of the robotic arm; based on the envelope sphere model and the motion trajectory of the robotic arm, collision detection is performed using a collision detection algorithm; and based on the collision detection results, the motion trajectory of the robotic arm is adjusted using a preset graded avoidance strategy.

[0050] In practice, by modeling the envelope spheres and simulating the motion trajectory of the robotic arm, the collision status between the envelope spheres is determined, thereby enabling collision judgment and adjustment of the robotic arm's motion trajectory.

[0051] The core of the envelope sphere model is to simplify the geometry of the links using spheres, enabling efficient collision detection calculations. Specifically, each robotic arm link (such as the upper arm, forearm, and wrist) corresponds to 1 to 3 spheres (determined based on the link length and cross-sectional dimensions). The diameter of the sphere is equal to the maximum cross-sectional dimension of the link (e.g., if the forearm diameter is 50mm, then the sphere diameter is 50mm). The centers of the spheres are evenly distributed along the central axis of the link, ensuring that the spheres completely enclose the link (with no exposed parts). For example, if the forearm length of the robotic arm is 200mm, an envelope sphere can be placed at each end and the midpoint, with the three spheres collectively covering the forearm, simplifying the complex link geometry into regular spheres.

[0052] The specific process of collision detection is as follows: Discretize the motion trajectory of each robotic arm by time step (e.g., 10ms / step), and calculate the real-time three-dimensional coordinates of each envelope sphere at each time step (based on the kinematic model of the robotic arm).

[0053] Check the sphere distances of all potential collision pairs, including "envelope spheres of different links of the same robotic arm", "envelope spheres of different robotic arms", and "envelope spheres and framework / pipeline models"; if the distance between the centers of the two spheres D < (sum of the radii of the two spheres) R1 + R2, then it is determined as a "potential collision".

[0054] Based on the physical parameters of the weld inspection device (such as probe lift-off distance ≥ 2mm), if the distance between the envelope ball and the frame surface is < 2mm, it is also considered a collision risk.

[0055] Possible collision detection results are: (1) No collisions: There are no potential collision pairs at all time steps, and the robot arm trajectory is directly usable; (2) Potential collision: Mark the time step of the collision and the link / component of the collision (e.g., "the forearm of robot arm 1 collided with the frame pipeline in step 50"), to provide a basis for adjustment of subsequent graded avoidance strategies (e.g., fine-tuning the joint angle of the robot arm and delaying the gantry movement sequence).

[0056] The robotic arm's trajectory is adjusted based on the collision detection results using a preset graded avoidance strategy. In this embodiment, the graded avoidance strategy can be set according to actual needs. Specifically, the graded avoidance strategy in this embodiment is as follows: (1) Low-risk avoidance: local joint fine-tuning (without changing the overall trajectory) When the collision risk exists only between a link of the robotic arm (such as the forearm) and a non-critical area of ​​the frame (such as a pipe near a non-weld), and the collision distance is close to the safety threshold (such as the difference in the distance between the envelope spheres < 1 mm), and there is no direct collision risk, only the angular acceleration / angular velocity of a single or two adjacent joints of the robotic arm is adjusted (the joint limit parameter constraints must be met), and the fine-tuning range is controlled within 5°; the core path of the probe scanning trajectory is not changed (to ensure the integrity of the weld coverage), and the distance between the envelope spheres and the interfering objects is increased only by fine-tuning the joints to maintain detection efficiency.

[0057] (2) Medium risk avoidance: staggered movement timing (adjusting the coordination rhythm of multiple components) When multiple robotic arms (such as a gantry robotic arm and a ground-rail robotic arm) have a risk of collision due to intersecting trajectories at the same time step, or when the robotic arm is too close to key interference objects in the structure (such as load-bearing pipes near welds) (the difference in the envelope sphere distance is <0.5mm), based on the total time constraint of the time optimization algorithm, the motion sequence of one of the robotic arms is adjusted (e.g., delaying the start of a certain trajectory by 0.5-2 seconds), or the local motion speed is reduced (not exceeding 30% of the original speed); the motion rhythm of the gantry / ground rail is synchronized and coordinated, for example, the gantry moves to a safe position first, and then the ground-rail robotic arm enters the intersection area to avoid multiple components occupying the same space at the same time; ensure that the total detection time after adjustment does not increase by more than 10% of the original plan (balancing efficiency and safety).

[0058] (3) High-risk avoidance: local trajectory reconstruction (replanning the scanning path) When collision risks cannot be avoided through fine-tuning or timing adjustments (e.g., direct overlap between the robotic arm link and the main structure, or envelope sphere distance < 0), or when collisions may damage the probe / distort the detection data (e.g., interference between the probe and the weld area), the local probe scanning trajectory is regenerated based on the B-spline curve to bypass the interference area (ensuring that the replanned trajectory still covers the target weld segment without any missed detections); the robotic arm motion trajectory is partially reconstructed, which may involve adjusting the motion parameters of 3 or more joints, requiring re-verification of collision detection; if the reconstructed local trajectory still conflicts with other components, the position of the gantry / ground rail can be adjusted in conjunction (e.g., the ground rail is shifted 50-100mm) to ensure no global collision risk; after reconstruction, the preset objective function (total time, energy consumption, and sum of joint accelerations) needs to be re-introduced for optimization to avoid a significant decrease in detection efficiency due to trajectory adjustments.

[0059] (4) Emergency risk avoidance: triggering shutdown and manual intervention (extreme situations) When real-time collision detection detects that the robotic arm has entered a collision countdown (e.g., the distance between the envelope spheres is <0.2mm), or when sensor malfunctions cause the collision risk to be unavoidable through algorithm adjustment, the robotic arm is immediately triggered to brake (joint lock), stopping the movement of the gantry / railway; an alarm signal is sent to the industrial control computer, prompting manual verification of the 3D model of the collision area against the actual framework structure (excluding misjudgments caused by point cloud modeling errors); after manual confirmation of the cause of interference, the framework model is manually corrected or the position of the detection device is adjusted, and the path planning process is restarted.

[0060] After obtaining the final motion trajectory of the robotic arm, the motion paths of each component are determined by optimizing and adjusting the parameters of the probe scanning trajectory, the robotic arm motion trajectory, the gantry motion trajectory, the ground rail motion trajectory, and the weld inspection device through a preset objective function. The preset objective function is as follows: ; in, , , These are the weighting coefficients, Total time The total energy consumption of all components. This represents the sum of joint accelerations.

[0061] It should be noted that the relationship between the preset objective function and each element is as follows:

[0062] The final optimization result achieves an optimal motion path that balances multiple objectives while satisfying the requirements of full weld coverage and hardware parameter constraints. The path planning method in this embodiment minimizes... This can reduce the total inspection time and improve inspection efficiency (for example, a traditional single robotic arm takes 2 hours, while optimized multi-robotic arm collaboration only takes 55 minutes); by minimizing This can reduce the drive energy consumption of robotic arms and gantry cranes (reducing total energy consumption by 20%~30%); by minimizing This allows the robotic arm to move more smoothly and avoids fluctuations in the probe lift-off distance caused by vibration (ensuring the stability of the detection signal).

[0063] S103: Determine the real-time paint thickness information of the structure corresponding to the probe, and dynamically determine the target defect detection frequency based on the real-time paint thickness information. At the same time, determine the probe lift-off distance through the distance control algorithm to ensure that the probe and the structure surface maintain a constant distance.

[0064] This embodiment dynamically adjusts the eddy current detection frequency by real-time detection of the paint layer thickness in the corresponding detection area, thereby greatly avoiding the impact of uneven paint layer thickness on defect detection. Furthermore, in practical implementation, an electromagnetic levitation device, based on a distance control algorithm, maintains a preset distance between the probe and the surface.

[0065] Specifically, the distance control algorithm is as follows: ; ; in, Electromagnetic output force, For proportional gain, For integral gain, For differential gain, For distance error, This is the ideal lift-off value (i.e., the distance between the probe and the surface of the frame). These are real-time measurements.

[0066] The process of dynamically determining the target defect detection frequency based on real-time paint layer thickness information in this embodiment is as follows: The probe integrates an electromagnetic induction paint thickness sensor, which simultaneously acquires the paint thickness of the corresponding area while scanning the weld. (The sampling frequency should be matched with the probe scanning speed, such as 10 times / second); A mapping table between paint layer thickness and detection frequency was pre-calibrated through experiments (e.g., ...). =50μm corresponding =20kHz, =100μm corresponding =35kHz), thick paint layers require higher frequencies to penetrate, while thin paint layers use lower frequencies to avoid noise; The industrial control computer receives the paint layer thickness data d in real time, outputs the target frequency command according to the mapping table, controls the excitation frequency of the eddy current probe, and realizes closed-loop feedback between paint layer thickness and detection frequency (response time < 50ms).

[0067] In eddy current testing, the paint layer attenuates the detection signal. If a low frequency is used on a thick paint layer, the signal attenuation is severe and the defect signal is masked. If a high frequency is used on a thin paint layer, electromagnetic noise is easily introduced, leading to misjudgment.

[0068] Traditional methods (fixed frequency) cannot handle uneven paint thickness (e.g., paint thickness differences of 30%~50% in different areas of the frame), requiring dynamic frequency matching to ensure signal-to-noise ratio. This embodiment solves the problem of paint interference with eddy current detection signals by determining the detection frequency based on real-time paint thickness. This eliminates the need for paint removal, polishing, and touch-up processes in traditional magnetic particle testing, saving 2~3 hours per unit and reducing material costs. It also has strong anti-interference capabilities, adapts to paint thicknesses of different levels, and improves defect identification accuracy to over 95% (compared to approximately 80% for traditional fixed frequency methods). Furthermore, it has wide adaptability, suitable for both new and old frame inspections (uneven paint wear on old frames and stable paint thickness on new frames).

[0069] S104: Control the probe to detect the frame according to the target defect detection frequency and probe lift-off distance, determine the defect information, and determine the weld defect information according to the defect information.

[0070] In this embodiment, the obtained defect information is processed by wavelet packet transform algorithm and adaptive noise cancellation algorithm to remove interference signals and determine the target signal; the impedance plane coordinate set is determined according to the target signal, and the area of ​​the target ellipse is determined by impedance plane ellipse fitting formula; the crack depth is determined by crack depth formula according to the area of ​​the target ellipse; the weld defect information in this embodiment includes at least crack depth.

[0071] The specific formula for the wavelet packet transform algorithm is as follows: ; in, The voltage signal acquired by the eddy current probe. The energy of a signal in a specific frequency band. At the frequency resolution level, For time location index, The wavelet packet is the basis function, derived from the mother wavelet. After scaling Peaceful relocation get.

[0072] The specific formula for the adaptive noise cancellation algorithm is as follows: ; ; ; in, This is a noisy defect signal. It is a pure noise signal. Defective models exist. Here, M represents the weighting coefficients, and M represents the filter order. Step size factor For time indexing, This is a tap index.

[0073] This embodiment determines the area of ​​the target ellipse using an impedance plane ellipse fitting formula, which is as follows: ; in, Let (h, k) be the coordinates of the impedance plane, (h, k) be the coordinates of the center of the ellipse, and a and b be the major and minor axes of the ellipse. The rotation angle is the direction of the principal axis of the ellipse.

[0074] Based on the area of ​​the ellipse, according to the formula... The area of ​​the target ellipse can be calculated by substituting the values ​​of a (where a is the semi-length of the major axis and b is the semi-length of the minor axis) into the fitted values ​​of a and b.

[0075] The crack depth is determined based on the area of ​​the target ellipse using the crack depth formula; the specific crack depth formula is as follows: ; Where K is the correction factor. To calculate the area of ​​the ellipse, The area of ​​the calibrated defect ellipse.

[0076] In this embodiment, the target signal is the eddy current detection signal after wavelet packet transform and adaptive noise cancellation. Essentially, it represents the impedance change of the probe coil (the eddy current effect causes a change in coil resistance). and reactance The specific process from the target signal to the impedance plane coordinate set (change) is as follows: (1) Signal decomposition: The target signal (voltage signal, which is linearly related to impedance change) is decomposed into voltage corresponding to resistance component through the signal conditioning circuit. Voltage corresponding to reactance component ; (2) Coordinate transformation: Based on the calibration parameters of the probe coil (e.g.) Conversion factor with R , Conversion factor with X ), calculate the real-time impedance value: , ; (3) Coordinate set generation: According to the sampling time sequence (e.g., sampling once every 1ms), record multiple sets of \((R,X)\) data to form an impedance plane coordinate set {( ),( ),…,( )}.

[0077] From the coordinate set to the target ellipse area, based on the impedance plane ellipse fitting formula, the specific details are as follows: For coordinate set {( ),…,( (In the formula, (X,Y)) is the impedance coordinate (R,X)). The ellipse parameters are fitted using the aforementioned impedance plane ellipse fitting formula: center (h,k), major and minor axes a / b, and rotation angle θ, to ensure that the deviation between the fitted ellipse and the coordinate points is minimized (the sum of squared residuals is minimized). Then, based on the fitted a and b, the area of ​​the target ellipse is calculated, and this area is used to calculate the crack depth.

[0078] This embodiment accurately determines the frame model and weld model using frame point cloud information, and determines the motion path of each component based on the frame model and weld model respectively. This enables multiple robotic arms, gantry trusses, and ground rail platforms to work together to perform synchronous and comprehensive scanning of multiple areas of the frame, improving the efficiency of weld inspection. Furthermore, when the robotic arm's eddy current probe is scanning for defects, it adjusts the detection frequency and lift-off distance in real time based on the real-time information of the paint layer thickness. Combined with frequency domain filtering algorithms, it suppresses paint layer interference, thereby achieving high-precision defect identification without the need for paint removal. This avoids the signal distortion caused by the probe lift-off effect being easily affected by paint layer thickness fluctuations in traditional eddy current detection.

[0079] Example 2 In one or more embodiments, a bogie frame weld inspection device is disclosed, including a ground rail platform, a gantry truss disposed above the ground rail platform, a multi-degree-of-freedom robotic arm disposed on the ground rail platform and the gantry truss respectively, and a probe disposed on the robotic arm. The bogie frame weld inspection device controls the gantry truss, the ground rail platform, the multi-degree-of-freedom robotic arm and the probe to work together to inspect the welds of the target frame according to the above-described bogie frame weld inspection method.

[0080] In this embodiment, the frame weld inspection device also includes an industrial control computer, an intelligent inspection instrument, and a 3D industrial camera. Each component communicates with the industrial control computer through an integrated control network to realize model construction, path planning, and collaborative inspection of the bogie frame welds.

[0081] Furthermore, as an example and not a limitation, in some alternative embodiments, the eddy current probe employs an adaptive structure, with its front-end detection head incorporating a paint layer thickness sensing module. The probe body is fixed to the end of the robotic arm via a flexible connector, enabling it to automatically adjust the detection angle and lifting distance based on the undulations of the paint layer on the frame surface. The intelligent detector integrates signal excitation, data acquisition, and defect analysis modules, suppressing paint layer interference signals through a frequency domain filtering algorithm and outputting defect characteristic parameters in real time. A gantry truss spans the frame inspection area, with dual robotic arms installed at the bottom of the truss. The robotic arms are equipped with eddy current probes and six-dimensional force sensors at their ends for synchronous scanning of the weld seams on the upper surface of the frame. A ground-rail motion platform is laid parallel to the length of the frame, with dual robotic arms mounted on its surface and equipped with rotary joints. The probes are driven to move along the weld seam trajectory on the lower surface through a path interpolation algorithm. The robotic arm movement paths of the gantry and the ground rail are uniformly planned by the industrial control computer based on the frame weld seam model generated by a 3D industrial camera, and interference between multiple robotic arms is avoided through an avoidance algorithm. A 3D industrial camera is fixed at the center of the top of the gantry truss. Its binocular vision module is tilted and aligned with the structure to be inspected, and a 3D contour model of the weld is constructed using structured light projection and point cloud stitching technology. Both the gantry truss and the ground rail motion platform are equipped with high-precision servo motors and limit sensors. Anti-collision buffer devices are installed on both sides of the gantry beam, and laser positioning marks are laid on the surface of the ground rail. Infrared distance sensors are installed on the base of each robotic arm and the end effector. When the distance between the robotic arms is less than the safety threshold, the industrial control computer triggers an emergency braking protocol and replans the path.

[0082] Example 3 In one or more embodiments, a bogie frame weld inspection system is disclosed, combined with Figure 2 Specifically, it includes: The model determination module is configured to acquire point cloud information of the steering frame and determine the frame model and weld model through the point cloud information. The trajectory determination module is configured to determine the motion paths of each component of the weld detection device based on the framework model and the weld model, wherein the weld detection device includes at least a probe and a robotic arm; The detection parameter dynamic adjustment module is configured to determine the real-time paint layer thickness information of the structure corresponding to the probe, and dynamically determine the target defect detection frequency based on the real-time paint layer thickness information. At the same time, the probe lift-off distance is determined through a distance control algorithm to ensure that the probe and the structure surface maintain a constant distance. The defect information determination module is configured to control the probe to detect the frame according to the target defect detection frequency and the probe lift-off distance, determine the defect information, and determine the weld defect information according to the defect information.

[0083] It should be noted that the specific implementation methods of the above modules are exactly the same as those in Example 1, and will not be described in detail again.

[0084] Example 4 In one or more embodiments, a terminal device is disclosed, combined with Figure 3 It includes a processor and a memory. The processor stores a computer program for implementing instructions. The memory stores multiple instructions that are adapted to be loaded and executed by the processor for the bogie frame weld inspection method described in Embodiment 1.

[0085] It should be understood that in this embodiment, the processor can be a central processing unit (CPU), or it can be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or any conventional processor, etc.

[0086] Memory may include read-only memory and random access memory, and provides instructions and data to the processor. A portion of memory may also include non-volatile random access memory. For example, memory may also store information about the device type.

[0087] In the implementation process, each step of the above method can be completed by the integrated logic circuits in the processor hardware or by software instructions.

[0088] While the specific embodiments of the present invention have been described above in conjunction with the accompanying drawings, this is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art without creative effort based on the technical solutions of the present invention are still within the scope of protection of the present invention.

Claims

1. A method for inspecting weld seams in a bogie frame, characterized in that, include: Obtain the point cloud information of the steering frame, and determine the frame model and weld model using the point cloud information; The motion paths of each component of the weld inspection device are determined based on the framework model and the weld model. Specifically, the spatial distribution of interference objects on the framework model is determined, and the probe scanning trajectory is determined by combining the weld model and the spatial distribution of interference objects using B-spline curves. The robotic arm's motion trajectory is determined by inverse kinematics based on the probe scanning trajectory and time optimization algorithm. The gantry movement trajectory and the ground track movement trajectory are determined by the movement trajectory of the robotic arm; After determining the robotic arm's motion trajectory, the process further includes: using the probe scanning trajectory, robotic arm motion trajectory, gantry motion trajectory, ground rail motion trajectory, and weld seam scanning parameters as constraints, optimizing and adjusting the trajectory using a preset objective function to determine the final motion path of each component; the preset objective function is: ； in, , , These are the weighting coefficients, Total time The total energy consumption of all components. This represents the sum of joint accelerations. The weld inspection device includes at least a probe and a robotic arm; The real-time paint thickness information of the structure in the area corresponding to the probe is determined, and the target defect detection frequency is dynamically determined based on the real-time paint thickness information. At the same time, the probe lift-off distance is determined through a distance control algorithm to ensure that the probe and the structure surface maintain a constant distance. The probe is controlled to detect the frame according to the target defect detection frequency and probe lift-off distance to determine defect information, and weld defect information is determined according to the defect information.

2. The method for inspecting weld seams in a bogie frame as described in claim 1, characterized in that, Obtain the point cloud information of the steering frame, and determine the frame model and weld model using the point cloud information, specifically as follows: An architecture model is constructed based on the point cloud information, and the architecture model is matched with a standard architecture model in a preset database to determine the matching degree; the preset database stores standard models of various types of steering architectures and corresponding weld seam models; When the matching degree is greater than the preset matching value, the target architecture model is determined, and then the weld model corresponding to the target architecture model is determined.

3. The method for inspecting weld seams in a bogie frame as described in claim 1, characterized in that, The time optimization algorithm minimizes the total motion time for the robotic arm to complete the probe scanning trajectory by adjusting the motion timing and velocity curves of each joint of the robotic arm, under the constraint that the angular acceleration and angular velocity of each joint of the robotic arm do not exceed the corresponding limit parameters of the robotic arm.

4. The method for inspecting weld seams in a bogie frame as described in claim 1, characterized in that, After determining the robotic arm's motion trajectory, the following steps are also included: Construct an envelope sphere model for each link of the robotic arm; The robotic arm's motion trajectory is discretized by time step, and the real-time three-dimensional coordinates of each envelope sphere at each time step are calculated. All sphere pairs with potential collisions are identified by calculating the distance between the spheres, and all spheres that do not meet the detection distance parameters are identified by calculating the distance between the spheres and the structure surface. The final collision detection result is then determined. Based on the detection results, the movement trajectory of the robotic arm is adjusted according to a preset graded avoidance strategy.

5. The method for inspecting weld seams in a bogie frame as described in claim 1, characterized in that, The target defect detection frequency is dynamically determined based on the real-time paint layer thickness information, specifically as follows: A pre-calibrated table showing the correspondence between paint layer thickness and detection frequency is used to determine the corresponding detection frequency based on the real-time paint layer thickness.

6. The method for inspecting weld seams in a bogie frame as described in claim 1, characterized in that, The specific details of determining weld defect information based on the aforementioned defect information are as follows: The weld defect information includes at least the crack depth; The defect information is processed using wavelet packet transform algorithm and adaptive noise cancellation algorithm to determine the target signal; The impedance plane coordinate set is determined based on the target signal, and the area of ​​the target ellipse is determined based on the impedance plane coordinate set using the impedance plane ellipse fitting formula. The crack depth is determined using the crack depth formula based on the area of ​​the target ellipse.

7. A bogie frame weld inspection system, employing a bogie frame weld inspection method as described in any one of claims 1-6, characterized in that, include: The model determination module is configured to acquire point cloud information of the steering frame and determine the frame model and weld model through the point cloud information. The trajectory determination module is configured to determine the motion paths of each component of the weld detection device based on the framework model and the weld model, wherein the weld detection device includes at least a probe and a robotic arm; The detection parameter dynamic adjustment module is configured to determine the real-time paint layer thickness information of the structure corresponding to the probe, and dynamically determine the target defect detection frequency based on the real-time paint layer thickness information. At the same time, the probe lift-off distance is determined through a distance control algorithm to ensure that the probe and the structure surface maintain a constant distance. The defect information determination module is configured to control the probe to detect the frame according to the target defect detection frequency and the probe lift-off distance, determine the defect information, and determine the weld defect information according to the defect information.

8. A terminal device comprising a processor and a memory, the processor for implementing instructions; the memory for storing multiple instructions, characterized in that, The instructions are adapted to be loaded by a processor and executed as the bogie frame weld inspection method according to any one of claims 1-6.

Images