A TIG welding molten pool collapse risk control method based on a galvanometer and space-time characteristics cooperation

Abstract

The present application relates to a kind of TIG welding molten pool collapse risk control method based on cooperation of galvanometer and space-time characteristics, belong to TIG welding process monitoring and control technical field.The method is by single galvanometer changes the projection position of line structured light on molten pool surface, according to matching time sequence, laser stripe image is collected and initial three-dimensional point cloud is reconstructed by visual camera with optical filter;According to robot end pose data, coordinate compensation is carried out, and molten pool three-dimensional point cloud in unified coordinate system is obtained;According to the position variation of corresponding space sampling position along gravity direction at adjacent sampling time, collapse change rate is calculated, and space-time characteristic data set is formed with three-dimensional coordinates;Risk value of molten pool collapse is obtained by fusion processing, according to the comparison result of the risk value and hierarchical control threshold, adjust galvanometer scanning parameter, image acquisition frequency and welding parameter.The present application can improve the real-time and stability of molten pool collapse risk monitoring and control, reduce the risk of molten pool excessive collapse and burn-through.

Description

Technical Field

[0001] This invention relates to a method for controlling the risk of TIG welding pool collapse based on the synergy of galvanometer and spatiotemporal characteristics, belonging to the field of TIG welding process monitoring and closed-loop control technology. Background Technology

[0002] In recent years, TIG welding has been widely used in thin plate and high-quality welding applications due to its advantages such as stable arc, good weld formation quality, and ease of control. Insufficient heat input can easily lead to defects such as incomplete penetration and lack of fusion; excessive heat input can easily cause the molten pool to collapse or even burn through. The penetration state of TIG welding is usually judged based on information such as welding current, arc voltage, molten pool image, or temperature field. Visual inspection has the advantages of being non-contact and providing rich information, and has been used for detecting molten pool geometry and weld misalignment. Existing visual inspection methods mostly focus on single-frame images or two-dimensional features, making it difficult to reflect the three-dimensional morphology of the molten pool surface and its collapse trend. Welding arc light, fumes, and spatter can also affect image acquisition quality. During robotic welding, the movement and posture changes of the welding torch can cause coordinate shifts in the molten pool morphology data, affecting the judgment of the penetration state. Some visual acquisition methods have fixed observation ranges and sampling frequencies, making it difficult to adjust the sampling area according to local risks, and also lack hierarchical closed-loop control coordinated with welding equipment. This invention uses galvanometer scanning, robot pose compensation, and spatiotemporal feature fusion to determine and classify the risk of molten pool collapse, thereby improving the real-time performance and stability of molten pool collapse risk monitoring and control. Summary of the Invention

[0003] A method for controlling the risk of TIG weld pool collapse based on the synergy of galvanometer and spatiotemporal features is characterized by the following: the method is implemented by a system including a vision acquisition device, a welding robot and supporting welding equipment, a data processing unit, and a closed-loop control unit; the vision acquisition device includes a vision camera with a filter, a line laser, a protective glass, a single galvanometer, and a galvanometer control device; the galvanometer control device drives the single galvanometer to deflect, thereby changing the projection position of the line structured light projected by the line laser on the surface of the weld pool, and causes the vision camera to acquire the laser stripe image formed on the surface of the weld pool according to a timing sequence matching the scanning of the single galvanometer; based on a pre-calibrated line structured light vision measurement model, the laser stripe image is reconstructed in three dimensions to obtain an initial three-dimensional point cloud on the surface of the weld pool; the data processing unit performs coordinate compensation on the initial three-dimensional point cloud according to the end pose data fed back by the welding robot. The process involves several steps: First, a three-dimensional point cloud of the molten pool is obtained in a unified coordinate system. Then, the corresponding spatial sampling positions between adjacent sampling times are determined based on the molten pool's three-dimensional point cloud. The positional change of these corresponding spatial sampling positions along the direction of gravity is calculated, and the collapse rate is obtained based on this positional change. The three-dimensional coordinates of each spatial sampling position and the corresponding collapse rate are combined during continuous sampling to obtain a spatiotemporal feature dataset. The data processing unit performs fusion processing on the spatiotemporal feature dataset to obtain a molten pool collapse risk value, which characterizes molten pool collapse anomalies and burn-through risks. The closed-loop control unit generates a graded closed-loop control command based on the comparison result between the molten pool collapse risk value and a preset graded control threshold. The visual acquisition device and the supporting welding equipment are adjusted according to the graded closed-loop control command to achieve closed-loop control of TIG welding molten pool collapse risk.

[0004] A method for controlling the risk of TIG weld pool collapse based on the synergy of galvanometer and spatiotemporal features is characterized in that: the visual acquisition device is set in front of the welding direction, and its detection range covers the weld pool area; the filter is installed at the front end of the lens of the visual camera, and the transmission band of the filter matches the emission wavelength of the line laser to reduce the influence of welding arc light and ambient light on the acquisition of laser stripe images; the projection optical path of the line laser and the imaging optical path of the visual camera are both reflected by the single galvanometer; the correspondence between the deflection angle of the single galvanometer, the projection position of the line structured light, and the imaging area of ​​the visual camera is obtained through pre-calibration; the protective glass is set on the light-emitting side of the visual acquisition device to isolate welding spatter and fumes.

[0005] A method for controlling the risk of TIG weld pool collapse based on the synergy of galvanometer and spatiotemporal features is characterized by the following: the coordinate compensation includes: transforming the initial 3D point cloud to a unified local welding coordinate system based on the coordinate transformation relationship between the pre-calibrated vision acquisition device and the robot end-effector coordinate system and the end-effector pose data fed back by the welding robot; determining the welding torch pose change based on the end-effector pose data of consecutive adjacent frames, and obtaining the pose inverse transformation matrix corresponding to the pose change through homogeneous coordinate transformation. ; through the inverse pose transformation matrix The transformed 3D point cloud is subjected to reverse compensation to reduce the point cloud coordinate offset caused by welding torch movement or attitude changes, resulting in a 3D point cloud of the molten pool in a unified coordinate system. The construction of the spatiotemporal feature dataset includes: determining the corresponding spatial sampling positions between adjacent sampling times in the compensated molten pool 3D point cloud, and extracting the first... The position change along the direction of gravity at each corresponding spatial sampling position is calculated using the discrete-time derivative equation based on the position change and the time interval between adjacent sampling times. The rate of collapse change at each corresponding spatial sampling location The rate of collapse change at each corresponding spatial sampling location; The spatiotemporal feature dataset is obtained by combining the corresponding three-dimensional coordinates according to the sampling time.

[0006] A method for controlling the risk of TIG weld pool collapse based on the synergy of galvanometer and spatiotemporal features, characterized in that: the fusion processing includes: calculating the spatial curvature gradient of the spatial sampling position on the weld pool surface based on the three-dimensional coordinates in the spatiotemporal feature dataset. The spatial distribution variance is calculated based on the collapse rate in the spatiotemporal feature dataset. The maximum collapse depth of the molten pool surface was determined based on the compensated 3D point cloud of the molten pool. and the proportion of high-risk areas ; the maximum collapse depth of the molten pool surface Spatial distribution variance of the rate of change of collapse Spatial curvature gradient and the proportion of high-risk areas Normalized weighted fusion is performed to obtain the risk value of molten pool collapse. The risk value of the molten pool collapse Used to characterize abnormal molten pool collapse and the degree of burn-through risk.

[0007] A method for controlling the risk of TIG weld pool collapse based on the synergy of galvanometer and spatiotemporal characteristics, characterized in that: the hierarchical closed-loop response executed by the closed-loop control unit according to the hierarchical closed-loop control command includes: the preset hierarchical control threshold includes an early warning threshold. Limit threshold and stop welding threshold ,and Before welding begins, set the initial welding current according to the current welding conditions. When continuous Risk value of melt pool collapse at each sampling time satisfy At this time, the system enters the welding adjustment state; in the welding adjustment state, high-risk areas on the surface of the molten pool are determined according to the spatiotemporal feature dataset, and the single galvanometer is controlled to reduce the scanning range so that the reduced scanning range covers the high-risk areas; the reduced local scanning amplitude satisfy In the formula, To set the scanning amplitude, Let be the amplitude compression coefficient, and Simultaneously, the scanning frequency of the single galvanometer for the high-risk area is increased, and the image acquisition frequency of the vision camera is correspondingly increased; when continuously Risk value of melt pool collapse at each sampling time satisfy At the same time, the welding power supply in the supporting welding equipment is controlled to perform flexible current reduction, lowering the current to the initial welding current. of times, of which ,and When continuous Risk value of melt pool collapse at each sampling time satisfy At that time, the welding power supply is controlled to perform rigid current reduction, lowering the current to the initial welding current. of times, of which The proportionality coefficient and It is set to ensure that the output current after both flexible and rigid current reduction is within the allowable current range corresponding to the current welding condition; when the system is in welding adjustment state and continuously... Risk value of melt pool collapse at each sampling time satisfy At the same time, the single galvanometer is controlled to return to the set scanning range and set scanning frequency, the vision camera is controlled to return to the set image acquisition frequency, and the welding power supply is controlled to return to the initial welding current at a preset recovery rate. The risk value of melt pool collapse at any sampling time satisfy At that time, the supporting welding equipment is controlled to perform welding stop protection. Among them, , , and The number of consecutive judgments is preset, and all values ​​are positive integers.

[0008] Beneficial effects of the invention

[0009] This invention addresses the challenges of timely identification of molten pool collapse trends during TIG welding, the tendency of welding torch movement and posture changes to cause coordinate shifts in the molten pool's three-dimensional morphology data, and the difficulty of adapting fixed sampling methods to local risk changes in the molten pool. It proposes a TIG welding molten pool collapse risk control method based on the synergy of galvanometer and spatiotemporal features.

[0010] This invention acquires laser stripe images of the molten pool surface through line structured light vision measurement and reconstructs an initial three-dimensional point cloud of the molten pool surface. Coordinate compensation is performed on the initial three-dimensional point cloud based on the end-effector pose data fed back by the welding robot to obtain a molten pool three-dimensional point cloud in a unified coordinate system. The collapse rate is calculated based on the position change along the gravity direction and time interval between adjacent sampling moments of the corresponding spatial sampling positions, and combined with the three-dimensional coordinates of the corresponding spatial sampling positions to form a spatiotemporal feature dataset. A molten pool collapse risk value is obtained through multi-feature fusion, and the scanning range and frequency of the single galvanometer, the image acquisition frequency of the vision camera, and the welding parameters of the supporting welding equipment are adjusted based on the comparison result between the risk value and a preset graded control threshold.

[0011] Compared with existing technologies, this invention can reduce the impact of welding torch movement and posture changes on the three-dimensional morphology detection results of the molten pool. It combines the spatial morphology of the molten pool surface and the rate of collapse change to judge the molten pool collapse anomaly and burn-through risk. Based on the risk level, it implements local scanning enhancement and graded adjustment of welding parameters, thereby improving the real-time performance and stability of TIG welding molten pool collapse risk monitoring and control, and reducing the probability of excessive molten pool collapse and burn-through. Attached Figure Description

[0012] Figure 1 This is a schematic diagram of the main structure of the TIG weld pool collapse risk control system of the present invention, wherein the arrow indicates the welding direction;

[0013] In the diagram, 1 is the vision acquisition device, 2 is the welding robot, 3 is the TIG welding torch, and 4 is the workpiece.

[0014] Figure 2 This is a schematic diagram of the structure and optical path of the visual acquisition device of the present invention, wherein the arrows indicate the direction of light propagation;

[0015] In the diagram, 5 is the vision camera, 6 is the line laser, 7 is the galvanometer control device, 8 is the single galvanometer, 9 is the protective glass, 10 is the imaging optical path of the vision camera, 11 is the line structured light projection optical path, and 12 is the filter.

[0016] Figure 3This is a flowchart of the TIG welding pool collapse risk control method based on the synergy of galvanometer and spatiotemporal characteristics according to the present invention. Detailed Implementation

[0017] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0018] Example 1:

[0019] like Figure 1 and Figure 2 As shown, a method for controlling the risk of TIG weld pool collapse based on the synergy of galvanometer and spatiotemporal features is implemented by a system including a vision acquisition device, a welding robot and supporting welding equipment, a data processing unit, and a closed-loop control unit. The supporting welding equipment includes a TIG welding torch, a welding power source, and a shielding gas supply device; in the case of filler wire TIG welding, it also includes a wire feeder. In one embodiment, the data processing unit and the closed-loop control unit can be integrated into the same industrial computer, or they can be implemented separately as an image processing board, an embedded processor, a motion controller, or a robot control cabinet. The closed-loop control unit controls the deflection of a single galvanometer through a galvanometer control device and adjusts the output parameters of the welding power source through a welding power source control interface.

[0020] like Figure 1 As shown, the welding robot 2 is equipped with a TIG welding torch 3, which moves relative to the workpiece 4 along the welding direction. A vision acquisition device 1 is positioned in front of the TIG welding torch 3 along the welding direction, and its detection range covers the molten pool area. A pre-calibrated relative pose relationship is maintained between the vision acquisition device 1 and the end effector of the welding robot 2.

[0021] like Figure 2As shown, the vision acquisition device includes a vision camera 5, a line laser 6, a galvanometer control device 7, a single galvanometer 8, a protective glass 9, and a filter 12 mounted on the front of the lens of the vision camera 5. The transmission wavelength of the filter 12 is matched with the emission wavelength of the line laser 6 to reduce the influence of welding arc light and ambient light on the laser stripe image. The imaging optical path 10 of the vision camera 5 and the projection optical path 11 of the line laser 6 are both reflected by the single galvanometer 8. The protective glass 9 is located on the light-emitting side of the vision acquisition device to isolate welding spatter and fumes. The galvanometer control device 7 drives the single galvanometer 8 to deflect, thereby changing the projection position of the line structured light on the surface of the molten pool. The correspondence between the deflection angle of the single galvanometer 8, the projection position of the line structured light, and the imaging area of ​​the vision camera 5 is obtained through pre-calibration. The vision camera 5 acquires laser stripe images according to a timing sequence matched with the scanning of the single galvanometer 8. After the single galvanometer reaches the set deflection angle, the vision camera is triggered to expose, and the laser stripe image is associated with the corresponding galvanometer deflection angle and sampling time, thereby reducing the positional error caused by the asynchronous movement of the galvanometer and image acquisition.

[0022] The data processing unit performs 3D reconstruction of the laser stripe image based on a pre-calibrated line structured light vision measurement model to obtain an initial 3D point cloud on the molten pool surface. The line structured light vision measurement model includes the intrinsic and extrinsic parameters of the vision camera, the equations of the line laser plane corresponding to different galvanometer deflection angles, and the coordinate transformation relationship between the vision acquisition device coordinate system and the robot end effector coordinate system. The data processing unit extracts the center line of the laser stripe and calculates the 3D coordinates of points on the molten pool surface based on the intersection of the camera imaging ray corresponding to the center pixel of the stripe and the line laser plane.

[0023] The welding robot feeds back its end-effector pose data corresponding to the image sampling time to the data processing unit via the robot control device. The data processing unit transforms the initial 3D point cloud to a unified welding local coordinate system based on the pre-calibrated coordinate transformation relationship between the vision acquisition device and the robot's end-effector coordinate system.

[0024] The data processing unit determines the welding torch pose change based on the end pose data of consecutive adjacent frames, and obtains the pose inverse transformation matrix corresponding to the pose change using the homogeneous coordinate transformation method. ; through the inverse pose transformation matrix Reverse compensation is performed on the 3D point cloud after conversion between adjacent sampling times to reduce the point cloud coordinate offset caused by welding torch translation or attitude change, so as to obtain the 3D point cloud of the molten pool in a unified coordinate system.

[0025] The data processing unit determines the corresponding spatial sampling positions between adjacent sampling times in the compensated 3D point cloud of the molten pool. These corresponding spatial sampling positions can be determined through grid indexing in a unified coordinate system, neighbor search, point cloud registration, or local surface interpolation. Spatial sampling positions whose spatial distance or matching confidence does not meet preset conditions are discarded as invalid data. In this embodiment, the direction along gravity is considered positive, and the rate of collapse change... It can be represented as:

[0026]

[0027] In the formula, For the current sampling time, the first... The corresponding spatial sampling positions have their position coordinates along the direction of gravity. This represents the position coordinates of the corresponding spatial sampling location along the direction of gravity at the previous sampling time. This represents the time interval between adjacent sampling moments. When the value is zero, it indicates that the location has a tendency to collapse along the direction of gravity. To reduce the impact of measurement noise on the discrete-time derivative, time filtering or local surface fitting can be performed on the position coordinates before calculation.

[0028] The data processing unit, according to the sampling time, calculates the three-dimensional coordinates of each corresponding spatial sampling position. With the corresponding rate of collapse The data are combined to obtain a spatiotemporal feature dataset. This dataset can be stored in the form of a point set, feature matrix, or temporal tensor, and is used to characterize the spatial morphology of the molten pool surface and its collapse trend over time.

[0029] Example 2:

[0030] like Figure 3 As shown, this embodiment, based on embodiment 1, performs fusion processing on the spatiotemporal feature dataset to obtain the melt pool collapse risk value. In this embodiment, the positive z-axis of the unified welding local coordinate system is set along the direction of gravity.

[0031] The data processing unit calculates the spatial curvature gradient of the molten pool surface at the spatial sampling location based on the three-dimensional spatial coordinates in the spatiotemporal feature dataset, and calculates the spatial distribution variance of the collapse rate based on the collapse rate in the spatiotemporal feature dataset. The spatial curvature gradient is used to characterize the degree of local morphological change on the molten pool surface, and the spatial distribution variance of the collapse rate is used to characterize the dispersion of the collapse rate at different spatial sampling locations on the molten pool surface.

[0032] The data processing unit determines the maximum collapse depth of the molten pool surface based on the compensated 3D point cloud of the molten pool. The maximum collapse depth of the molten pool surface can be expressed as:

[0033]

[0034] In the formula, The maximum collapse depth of the molten pool surface. For the first The coordinates of the reference position corresponding to the spatial sampling position along the direction of gravity can be determined based on the historical sampling point cloud, the initial molten pool surface point cloud, or the average position of multiple adjacent historical sampling times during the welding stabilization phase.

[0035] The data processing unit calculates the average rate of collapse change at the spatial sampling locations on the surface of the molten pool involved in the fusion process:

[0036]

[0037] And calculate the spatial distribution variance of the collapse rate of change:

[0038]

[0039] In the formula, This represents the number of spatial sampling locations on the surface of the molten pool that are used in the calculation.

[0040] The data processing unit fits the local curvature of the molten pool surface based on the local neighborhood point cloud of each spatial sampling position in the compensated molten pool 3D point cloud, and calculates the spatial curvature gradient features:

[0041]

[0042] In the formula, This represents the spatial curvature gradient characteristic. For the first Local curvature of a spatial sampling location on the surface of a molten pool. Let be the gradient vector of the local curvature in the spatial neighborhood of the molten pool surface. The magnitude of the gradient vector is given. The spatial curvature gradient feature is used to characterize the degree of local depressions, abrupt changes, or morphological discontinuities on the surface of the molten pool.

[0043] The data processing unit also identifies high-risk areas based on the rate of collapse and the depth of collapse. When the... A spatial sampling location is identified as a high-risk spatial sampling location if it meets at least one of the following conditions:

[0044] or

[0045] In the formula, To preset the collapse rate threshold, This is a preset collapse depth threshold.

[0046] When the spatial sampling positions within the effective detection area of ​​the molten pool are uniformly distributed or the local surface areas corresponding to each spatial sampling position are the same, the area ratio of the high-risk area is... It can be represented as:

[0047]

[0048] In the formula, This represents the percentage of high-risk areas. The number of spatial sampling locations on the surface of the high-risk molten pool. When the distribution of point cloud spatial sampling locations is uneven, the area proportion of the high-risk region is determined by the ratio of the sum of the local surface areas corresponding to the high-risk spatial sampling locations to the total area of ​​the effective detection area of ​​the molten pool.

[0049] The data processing unit determines the maximum collapse depth of the molten pool surface. Spatial distribution variance of the rate of change of collapse Spatial curvature gradient characteristics and the proportion of high-risk areas Normalized weighted fusion is performed to obtain the risk value of molten pool collapse. :

[0050]

[0051] In the formula, This represents the risk value for molten pool collapse. These are the weighting coefficients corresponding to the spatial distribution variance of the maximum collapse depth of the molten pool surface, the collapse rate of change, the spatial curvature gradient characteristics, and the area proportion of high-risk regions, respectively. , , , For the normalization function, a minimum-maximum normalization function based on preset upper and lower limits can be used to ensure that each normalized feature is in the same numerical range.

[0052] In one implementation, the weighting coefficients satisfy:

[0053]

[0054]

[0055] above , Normalization parameters and weighting coefficients can be pre-calibrated based on historical sampling data of welding materials, plate thickness, welding current, welding speed, and the stable state of the molten pool, or they can be set according to different welding conditions. Molten pool collapse risk value. It is used to characterize the degree of risk of the current molten pool collapsing or burning through.

[0056] Example 3:

[0057] This embodiment, based on Embodiments 1 and 2, explains how the closed-loop control unit adjusts the melt pool collapse risk value. The process of executing a hierarchical closed-loop response.

[0058] like Figure 3 As shown, before welding begins, the initial welding current is set based on the current welding conditions, including welding material, plate thickness, welding speed, and whether filler wire is used. Set the scanning amplitude Set the scanning frequency, image acquisition frequency, and allowable process current range. The closed-loop control unit will output the melt pool collapse risk value from the data processing unit. Compare with preset hierarchical control thresholds. Preset hierarchical control thresholds include warning thresholds. Limit threshold and stop welding threshold ,and When the risk value of the molten pool collapses Reaching or exceeding the warning threshold At that time, pause the reference position. Updates are made to prevent abnormal collapse data from being included in the reference location.

[0059] When the risk value of molten pool collapse satisfy When the system is not in the welding adjustment state, the closed-loop control unit maintains the current scanning parameters, image acquisition frequency, and welding parameters, and continues to perform closed-loop monitoring.

[0060] When continuous Risk value of melt pool collapse at each sampling time satisfy At this point, the system enters the welding adjustment state. The closed-loop control unit determines the high-risk area based on the spatial distribution of the high-risk space sampling positions, and compresses the scanning range of a single galvanometer to the high-risk area according to the pre-calibrated correspondence between the spatial position and the galvanometer deflection angle. The compressed local scanning amplitude... satisfy:

[0061]

[0062] In the formula, Let be the amplitude compression coefficient, and Simultaneously, the closed-loop control unit increases the scanning frequency of the single galvanometer in high-risk areas and correspondingly increases the image acquisition frequency of the vision camera, ensuring that the camera acquisition timing continues to match the galvanometer scanning timing. This increases the amount of effective sampling data in high-risk areas and improves the real-time performance and accuracy of molten pool collapse risk monitoring.

[0063] When continuous Risk value of melt pool collapse at each sampling time satisfy At that time, the closed-loop control unit controls the welding power supply to perform flexible current reduction, adjusting the target output current to:

[0064]

[0065] In the formula, ,and Proportionality coefficient The current reduction can be segmented based on the position of the risk value between the warning threshold and the limit threshold. Flexible current reduction refers to a relatively gentle reduction of the welding current based on the degree of molten pool collapse risk when the risk value reaches the warning threshold but has not yet reached the limit threshold. This reduces the heat input to the molten pool while ensuring the welding current remains within the allowable range of the current welding process. The reduction range of the flexible current reduction is smaller than that of the rigid current reduction, and the specific reduction ratio can be pre-calibrated based on the molten pool collapse risk value, welding material, plate thickness, and welding speed.

[0066] When continuous Risk value of melt pool collapse at each sampling time satisfy At that time, the closed-loop control unit controls the welding power supply to perform rigid current reduction, adjusting the target output current to:

[0067]

[0068] In the formula, Rigid current reduction uses a higher current reduction margin than flexible current reduction to quickly reduce the heat input to the molten pool. The output current after reduction is also kept within the allowable current range of the current welding condition. The specific reduction ratio can be pre-calibrated according to the molten pool collapse risk value, welding material, plate thickness and welding speed.

[0069] proportionality coefficient and The value of should ensure that the output current after reduction is within the allowable current range of the current process corresponding to the current welding condition.

[0070] When the system is in the welding adjustment state, and continuously Risk value of melt pool collapse at each sampling time satisfy At the same time, the closed-loop control unit controls the single galvanometer to return to the set scanning range and set scanning frequency, controls the vision camera to return to the set image acquisition frequency, and controls the welding power supply to gradually return to the initial welding current according to the preset recovery rate. This is to prevent sudden changes in output current from causing instability in the molten pool again. The preset recovery rate is the increase in welding current per unit time, or the current increment between adjacent control cycles; it is pre-calibrated based on the welding material, plate thickness, welding speed, and the allowable current variation range of the process.

[0071] in, , , and The preset number of consecutive judgments is a positive integer. This preset number of consecutive judgments can be the same or different, and is preset based on the image acquisition frequency, closed-loop control cycle, molten pool dynamic response speed, and the urgency of the response corresponding to different control states. (Number of consecutive judgments) This is used to determine whether the risk of molten pool collapse has reached a level requiring welding adjustment, in order to suppress false triggering of the welding adjustment state caused by instantaneous measurement fluctuations. Among the continuous sampling times used to determine when the system enters the welding adjustment state, sampling times that simultaneously satisfy the risk range corresponding to flexible current reduction can be included in the number of consecutive judgments. Simultaneously, the sampling time that satisfies the risk interval corresponding to the rigid current reduction can be included in the number of consecutive judgments. There is no need to restart the counting after the system enters the welding adjustment state. The number of consecutive judgments corresponding to the flexible current reduction. Used to suppress frequent adjustments caused by instantaneous measurement fluctuations; rigid current reduction corresponds to the number of consecutive judgments. Used to improve the response speed to higher collapse risks while suppressing false triggers; restores the number of consecutive judgments corresponding to the control. This is used to prevent repeated switching of control states when the risk value of molten pool collapse fluctuates around the warning threshold. When the risk value of molten pool collapse fluctuates repeatedly between two adjacent risk intervals, the closed-loop control unit maintains the current control state until the risk value of molten pool collapse continuously meets the judgment condition corresponding to another risk interval, and then switches to the corresponding control state.

[0072] The closed-loop control unit can also adjust welding parameters such as welding speed, arc length, or shielding gas flow rate within the allowable range of the process. In the case of filler wire TIG welding, the closed-loop control unit can also coordinate with the wire feeder to adjust the wire feed speed to assist in adjusting the thermal balance of the molten pool.

[0073] When the risk value of melt pool collapse at any sampling time satisfy At this time, the closed-loop control unit controls the supporting welding equipment to perform welding stop protection. The welding stop protection includes cutting off or rapidly reducing the welding current, stopping the robot from continuing to move along the welding path, maintaining the supply of shielding gas for a delay according to a preset sequence, and stopping wire feeding in the wire feeding condition.

[0074] in, , , , , , and The parameters can be predetermined based on the welding materials, plate thickness, and stable welding conditions, through process experiments, offline calibration, or empirical parameter tables.

[0075] This method sequentially performs the following steps: galvanometer collaborative acquisition of laser stripe images, initial 3D point cloud reconstruction, robot end-effector pose coordinate compensation, spatiotemporal feature dataset extraction, fusion calculation of molten pool collapse risk value, generation of hierarchical closed-loop control commands, and adjustment of the visual acquisition device and supporting welding equipment. After completing the control adjustment, it returns to the closed-loop monitoring process, thereby realizing continuous detection and closed-loop control of TIG weld molten pool collapse risk.

Claims

1. A method for controlling the risk of TIG weld pool collapse based on the synergy of galvanometer and spatiotemporal characteristics, characterized in that: The method is implemented by a system comprising a vision acquisition device, a welding robot and supporting welding equipment, a data processing unit, and a closed-loop control unit. The vision acquisition device includes a vision camera with a filter, a line laser, a protective glass, a single galvanometer, and a galvanometer control device. The galvanometer control device drives the single galvanometer to deflect, thereby changing the projection position of the line structured light projected by the line laser on the surface of the molten pool, and causes the vision camera to acquire the laser stripe image formed on the surface of the molten pool according to a timing sequence matching the scanning of the single galvanometer. Based on a pre-calibrated line structured light vision measurement model, the laser stripe image is reconstructed in three dimensions to obtain an initial three-dimensional point cloud on the surface of the molten pool. The data processing unit performs coordinate compensation on the initial three-dimensional point cloud according to the end-effector pose data fed back by the welding robot to obtain a three-dimensional point cloud of the molten pool in a unified coordinate system. The spatial sampling positions between adjacent sampling times are determined based on the 3D point cloud of the molten pool. The position change of the corresponding spatial sampling position along the direction of gravity is calculated, and the collapse rate is obtained based on the position change. The 3D coordinates of each spatial sampling position and the corresponding collapse rate are combined during continuous sampling to obtain a spatiotemporal feature dataset. The data processing unit performs fusion processing on the spatiotemporal feature dataset to obtain a molten pool collapse risk value, which is used to characterize molten pool collapse anomalies and burn-through risks. The closed-loop control unit generates a graded closed-loop control command based on the comparison result between the molten pool collapse risk value and a preset graded control threshold. The visual acquisition device and the supporting welding equipment are adjusted according to the graded closed-loop control command to achieve closed-loop control of TIG welding molten pool collapse risk.

2. The method for controlling the risk of TIG weld pool collapse based on the synergy of galvanometer and spatiotemporal characteristics as described in claim 1, characterized in that: The vision acquisition device is positioned in front of the welding direction, and its detection range covers the molten pool area. The filter is mounted on the front of the lens of the vision camera, and the transmission wavelength of the filter matches the emission wavelength of the line laser to reduce the influence of welding arc light and ambient light on the acquisition of laser stripe images. The projection optical path of the line laser and the imaging optical path of the vision camera are both reflected by the single galvanometer. The correspondence between the deflection angle of the single galvanometer, the projection position of the line structured light, and the imaging area of ​​the vision camera is obtained through pre-calibration. The protective glass is set on the light-emitting side of the vision acquisition device to isolate welding spatter and fumes.

3. The method for controlling the risk of TIG weld pool collapse based on the synergy of galvanometer and spatiotemporal characteristics as described in claim 1, characterized in that: The coordinate compensation includes: transforming the initial 3D point cloud to a unified welding local coordinate system based on the pre-calibrated coordinate transformation relationship between the vision acquisition device and the robot's end-effector coordinate system and the end-effector pose data fed back by the welding robot; determining the welding torch pose change based on the end-effector pose data of consecutive adjacent frames, and obtaining the pose inverse transformation matrix corresponding to the pose change using the homogeneous coordinate transformation method. ; through the inverse pose transformation matrix The transformed 3D point cloud is subjected to reverse compensation to reduce the point cloud coordinate offset caused by welding torch movement or attitude changes, resulting in a 3D point cloud of the molten pool in a unified coordinate system. The construction of the spatiotemporal feature dataset includes: determining the corresponding spatial sampling positions between adjacent sampling times in the compensated molten pool 3D point cloud, and extracting the first... The position change along the direction of gravity at each corresponding spatial sampling position is calculated using the discrete-time derivative equation based on the position change and the time interval between adjacent sampling times. The rate of collapse change at each corresponding spatial sampling location The rate of collapse change at each corresponding spatial sampling location; The spatiotemporal feature dataset is obtained by combining the corresponding three-dimensional coordinates according to the sampling time.

4. The method for controlling the risk of TIG weld pool collapse based on the synergy of galvanometer and spatiotemporal characteristics as described in claim 1, characterized in that: The fusion process includes: calculating the spatial curvature gradient of the spatial sampling position on the surface of the molten pool based on the three-dimensional coordinates in the spatiotemporal feature dataset. The spatial distribution variance is calculated based on the collapse rate in the spatiotemporal feature dataset. The maximum collapse depth of the molten pool surface is determined based on the compensated 3D point cloud of the molten pool. and the proportion of high-risk areas ; the maximum collapse depth of the molten pool surface Spatial distribution variance of the rate of change of collapse Spatial curvature gradient and the proportion of high-risk areas Normalized weighted fusion is performed to obtain the risk value of molten pool collapse. The risk value of the molten pool collapse Used to characterize abnormal molten pool collapse and the degree of burn-through risk.

5. The method for controlling the risk of TIG weld pool collapse based on the synergy of galvanometer and spatiotemporal characteristics as described in claim 1, characterized in that: The hierarchical closed-loop response executed by the closed-loop control unit according to the hierarchical closed-loop control command includes: the preset hierarchical control threshold includes a warning threshold. Limit threshold and stop welding threshold ,and Before welding begins, set the initial welding current according to the current welding conditions. When continuous Risk value of melt pool collapse at each sampling time satisfy At this time, the system enters the welding adjustment state; in the welding adjustment state, high-risk areas on the surface of the molten pool are determined according to the spatiotemporal feature dataset, and the single galvanometer is controlled to reduce the scanning range so that the reduced scanning range covers the high-risk areas; the reduced local scanning amplitude satisfy In the formula, To set the scanning amplitude, Let be the amplitude compression coefficient, and Simultaneously, the scanning frequency of the single galvanometer for the high-risk area is increased, and the image acquisition frequency of the vision camera is correspondingly increased; when continuously Risk value of melt pool collapse at each sampling time satisfy At the same time, the welding power supply in the supporting welding equipment is controlled to perform flexible current reduction, lowering the current to the initial welding current. of times, of which ,and When continuous Risk value of melt pool collapse at each sampling time satisfy At that time, the welding power supply is controlled to perform rigid current reduction, lowering the current to the initial welding current. of times, of which The proportionality coefficient and It is set to ensure that the output current after both flexible and rigid current reduction is within the allowable current range corresponding to the current welding condition; when the system is in welding adjustment state and continuously... Risk value of melt pool collapse at each sampling time satisfy At the same time, the single galvanometer is controlled to return to the set scanning range and set scanning frequency, the vision camera is controlled to return to the set image acquisition frequency, and the welding power supply is controlled to return to the initial welding current at a preset recovery rate. The risk value of melt pool collapse at any sampling time satisfy At that time, the supporting welding equipment is controlled to perform welding stop protection. Among them, , , and The number of consecutive judgments is preset, and all values ​​are positive integers.

Images