A welding track real-time correction method based on integrated visual guidance

Abstract

The application discloses a kind of based on integrated visual guidance's welding track real-time correction method, it is related to welding automation technical field, including receiving the structured light vision image of weld area in front of welding torch, and the arc electric signal of welding torch end, the arc electric signal includes welding current transient value and welding voltage transient value;Current three-dimensional track coordinates and current welding torch height coordinates under the welding torch base coordinate system are respectively obtained from the structured light vision image and the arc electric signal;The present application distinguishes gradual change state from mutation state by introducing weld state determination mechanism, and different correction strategies are used for the two states respectively, thereby solving the problem of excessive correction or correction lag caused by simultaneous correction of track deviation and height deviation in the prior art.

Description

Technical Field

[0001] This invention relates to the field of welding automation technology, and in particular to a method for real-time correction of welding trajectories based on integrated vision guidance. Background Technology

[0002] In the field of welding automation, weld seam tracking and trajectory correction have always been core technologies for ensuring welding quality. Traditional welding robots typically operate using pre-programmed trajectories. However, in actual welding processes, due to factors such as workpiece assembly errors, thermal deformation, and the geometric irregularities of the weld seam itself, the preset trajectory often deviates from the actual weld seam position.

[0003] In existing technologies, single-sensor weld seam tracking methods each have their limitations. Structured light vision has high accuracy in horizontal trajectory recognition, but it is prone to failure under conditions of strong light interference or strong surface reflection; while arc sensing is sensitive to height changes and has a fast response speed, its accuracy in horizontal trajectory recognition is limited. Neither of these methods, when used alone, can meet the welding requirements under complex working conditions.

[0004] More critically, existing technologies lack a state-aware mechanism for correcting trajectory and height deviations, typically employing a unified strategy of correcting both simultaneously. When the weld is in a gradually changing state, the slow change in weld geometry causes trajectory and height deviations to exhibit coordinated drift characteristics. In this case, the essence of trajectory deviation is not that the welding torch deviates from the correct position, but rather that the weld centerline itself has objectively drifted. Forcibly correcting trajectory deviation causes the correction system to fight against the thermal deformation trend, resulting in continuous horizontal oscillations. When the weld is in a sudden change state, the weld geometry changes drastically in the horizontal direction, requiring a rapid response from the trajectory deviation. However, at this time, the molten pool surface experiences violent oscillations due to the sudden change in heat input, and the height deviation is mixed with a large amount of noise. Simultaneously, correcting the height causes the welding torch to shake up and down following the molten pool surface, introducing vertical instability. Existing technologies fail to distinguish between these two distinct operating conditions, resulting in horizontal oscillations at the point of gradual change and vertical shaking at the point of sudden change. These two defects alternate, severely restricting the stability of welding quality.

[0005] Therefore, there is a need for a real-time welding trajectory correction method that can integrate multi-source sensor information, possess weld condition perception capabilities, and implement corresponding correction strategies for different working conditions. Summary of the Invention

[0006] The purpose of this invention is to provide a real-time correction method for welding trajectories based on integrated vision guidance, so as to solve the problem of unstable welding quality caused by the limitations of a single sensor and the simplification of correction strategies in the prior art.

[0007] To achieve the above objectives, the present invention provides the following technical solution: A real-time welding trajectory correction method based on integrated vision guidance includes: Receive structured light visual images of the weld area in front of the welding torch, as well as electric arc signals from the end of the welding torch, wherein the electric arc signals include transient values ​​of welding current and transient values ​​of welding voltage; The current three-dimensional trajectory coordinates and current welding torch height coordinates in the welding torch base coordinate system are obtained from the structured light visual image and the arc electrical signal, respectively. Calculate the trajectory deviation vector between the current three-dimensional trajectory coordinates and the three-dimensional trajectory coordinates of the previous cycle, and the height deviation vector between the current welding torch height coordinates and the welding torch height coordinates of the previous cycle; It is determined whether the trajectory deviation vector and the height deviation vector have the same trend in time, and whether the fluctuation of the trajectory deviation vector is periodic or abrupt; the abruptness refers to the rate of change of the amplitude of the trajectory deviation vector within a single control cycle exceeding a preset abruptness threshold; when the trends are in the same direction and the fluctuation is periodic, the current weld is determined to be in a gradual change state; when the trends are not in the same direction or the fluctuation is abrupt, the current weld is determined to be in an abrupt state. When in a state of gradual change, a first comprehensive trajectory correction amount is generated from the current three-dimensional trajectory coordinates and the current welding torch height coordinates. The first comprehensive trajectory correction amount only corrects the height deviation vector, while the trajectory deviation vector remains unchanged at its current value. When in a sudden change state, a second comprehensive trajectory correction amount is generated from the current three-dimensional trajectory coordinates and the current welding torch height coordinates. The second comprehensive trajectory correction amount only corrects the trajectory deviation vector, while the height deviation vector remains unchanged. The first comprehensive trajectory correction amount or the second comprehensive trajectory correction amount is fed forward to the motion controller of the welding robot to drive the end of the welding torch to perform six-degree-of-freedom trajectory compensation.

[0008] Preferably, determining whether the trajectory deviation vector and the height deviation vector have the same trend in time, and determining whether the fluctuation of the trajectory deviation vector is periodic or abrupt, includes: calculating the first direction of change of the trajectory deviation vector in three consecutive periods, and the second direction of change of the height deviation vector in the corresponding three periods; when both the first direction of change and the second direction of change are increasing or both are decreasing, the trend is determined to be in the same direction; calculating the number of fluctuation periods of the trajectory deviation vector in the last ten periods; when the number of fluctuation periods is greater than or equal to three, the fluctuation is determined to be periodic; when the number of fluctuation periods is less than three, if the rate of change of the amplitude of the trajectory deviation vector in a single control period exceeds a preset abrupt change threshold, the fluctuation is determined to be abrupt; if it does not exceed the abrupt change threshold, the fluctuation is determined to be a non-periodic disturbance and is processed as an abrupt change.

[0009] Preferably, when in a progressively changing state, generating a first comprehensive trajectory correction amount from the current three-dimensional trajectory coordinates and the current welding torch height coordinates includes: generating a height correction component from the current welding torch height coordinates and the height deviation vector; the trajectory deviation vector does not participate in the generation of the correction component; and using the height correction component as the first comprehensive trajectory correction amount.

[0010] Preferably, when in a sudden change state, generating a second comprehensive trajectory correction amount from the current three-dimensional trajectory coordinates and the current welding torch height coordinates includes: generating a trajectory correction component from the current three-dimensional trajectory coordinates and the trajectory deviation vector; the height deviation vector does not participate in the generation of the correction component; and using the trajectory correction component as the second comprehensive trajectory correction amount.

[0011] Preferably, the step of obtaining the current three-dimensional trajectory coordinates and current welding torch height coordinates in the welding torch base coordinate system from the structured light visual image and the arc electrical signal respectively includes: extracting the pixel coordinates of weld feature points from the structured light visual image and converting them into the current three-dimensional trajectory coordinates in the welding torch base coordinate system according to camera calibration parameters; calculating the transient value of arc length from the arc electrical signal includes: calculating the transient value of arc resistance according to the ratio of the transient value of welding voltage to the transient value of welding current; under preset welding process parameters, obtaining the transient value of arc length from the transient value of arc resistance by looking up the table according to the preset mapping relationship table of arc resistance and arc length; and converting the transient value of arc length into the current welding torch height coordinates in the welding torch base coordinate system according to the mapping relationship between the transient value of arc length and welding torch height.

[0012] Preferably, when in a gradual change state, keeping the trajectory deviation vector unchanged includes: using the trajectory deviation vector as a position drift reference for the welding torch tip relative to the weld centerline; using the position drift reference as the reference origin for trajectory compensation of the welding torch tip in subsequent cycles; and resetting the position drift reference to the current trajectory deviation vector when the gradual change state continues for more than a preset reference holding period or when the change trend is detected to be reversed.

[0013] Preferably, when in a sudden change state, keeping the height deviation vector unchanged includes: using the height deviation vector as a height oscillation reference for the welding torch tip relative to the current molten pool surface; using the height oscillation reference as the reference origin for height compensation of the welding torch tip in subsequent cycles; and resetting the height oscillation reference to the current height deviation vector when the sudden change state ends and transitions to a gradual change state.

[0014] Preferably, generating a height correction component from the current welding torch height coordinates and the height deviation vector includes: calculating a height correction weight based on the ratio of the amplitude of the height deviation vector to the amplitude of the trajectory deviation vector; setting the height correction weight to a preset maximum weight value when the amplitude of the trajectory deviation vector is zero; directly generating the height correction component from the current welding torch height coordinates and the height deviation vector when the amplitude of the trajectory deviation vector is not zero; and generating the height correction component from the current welding torch height coordinates and the height deviation vector after smoothing when the height correction weight is less than or equal to the weight threshold.

[0015] Preferably, the step of generating a trajectory correction component from the current three-dimensional trajectory coordinates and the trajectory deviation vector includes: calculating a trajectory correction weight based on the ratio of the amplitude of the trajectory deviation vector to the amplitude of the height deviation vector; setting the trajectory correction weight to a preset maximum weight value when the amplitude of the height deviation vector is zero; setting the trajectory correction weight to a preset maximum weight value when the amplitude of the height deviation vector is not zero; dividing the amplitude of the trajectory deviation vector by the amplitude of the height deviation vector when the trajectory correction weight is greater than a preset weight threshold; directly generating the trajectory correction component from the current three-dimensional trajectory coordinates and the trajectory deviation vector when the trajectory correction weight is less than or equal to the weight threshold; and generating the trajectory correction component from the current three-dimensional trajectory coordinates and the trajectory deviation vector after amplitude limiting processing, wherein the amplitude limiting processing refers to limiting the amplitude of the trajectory correction component to a range not exceeding a preset maximum correction amplitude, wherein the maximum correction amplitude is 0.5 mm to 1.0 mm.

[0016] Preferably, using the trajectory deviation vector as the position drift reference of the welding torch tip relative to the weld centerline includes: decomposing the trajectory deviation vector into a direction component and an amplitude component in the horizontal plane; the direction component serves as the position drift direction of the welding torch tip in the horizontal plane; the amplitude component serves as the position drift distance of the welding torch tip in the horizontal plane; the direction component maintains a constant cumulative value during the duration of the gradual change state; that is, the initial value of the direction component locked from the first control cycle of entering the gradual change state is no longer updated with local fluctuations of the weld in subsequent cycles until the state ends or a reset condition is triggered; the amplitude component is updated with the amplitude of the current trajectory deviation vector, that is, in each control cycle, the amplitude of the current trajectory deviation vector is used as the new amplitude component value, thereby reflecting the latest change in the drift distance of the weld centerline.

[0017] The technical solutions provided by the embodiments of this disclosure have at least the following beneficial effects: This invention achieves synchronous perception of weld trajectory and welding torch height by fusing two information sources: structured light visual images and electric arc signals, overcoming the failure risk of a single sensor under complex working conditions. By introducing a weld state determination mechanism, it distinguishes between gradual change states and abrupt change states, and adopts differentiated correction strategies for each state. In the gradual change state, it keeps the trajectory deviation constant and focuses on correcting the height to avoid horizontal oscillation; in the abrupt change state, it keeps the height deviation constant and focuses on correcting the trajectory to achieve rapid response. This solves the problem of overcorrection or correction lag caused by simultaneous correction of trajectory deviation and height deviation in the prior art. By using feedforward control to directly inject the correction amount into the motion controller, the correction delay is controlled within a single control cycle, achieving instantaneous response to geometrical changes in the weld. Attached Figure Description

[0018] Figure 1 This is a schematic diagram of the method structure of the present invention. Detailed Implementation

[0019] The technical solution of the present invention will be clearly and completely described below with reference to the embodiments. 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 of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0020] Please see Figure 1 This application provides a real-time correction method for welding trajectories based on integrated vision guidance, including: In the field of welding automation, weld seam tracking and trajectory correction have always been core technical aspects for ensuring welding quality. Traditional welding robots typically operate using pre-programmed trajectories. However, during actual welding, factors such as workpiece assembly errors, thermal deformation, and the geometric irregularities of the weld seam itself often cause deviations between the preset trajectory and the actual weld seam position. Existing single-sensor weld seam tracking methods each have limitations: structured light vision offers high accuracy in horizontal trajectory recognition, but it is prone to failure under conditions of strong light interference or intense surface reflection; while arc sensing is sensitive to height changes and has a fast response speed, its accuracy in horizontal trajectory recognition is limited. More critically, existing technologies lack a state-aware mechanism for correcting trajectory and height deviations, typically employing a unified strategy of correcting both simultaneously. This results in horizontal oscillations at gradual changes and vertical jitter at abrupt changes, with these two defects alternating, severely restricting the stability of welding quality.

[0021] Based on the aforementioned physical laws, this invention employs the following control logic: determining whether the trajectory deviation vector and the height deviation vector exhibit the same temporal trend, and determining whether the fluctuation of the trajectory deviation vector is periodic or abrupt; when the trends are the same and the fluctuation is periodic, the current weld is determined to be in a gradual change state; when the trends are different or the fluctuation is abrupt, the current weld is determined to be in an abrupt change state. This dual determination of the same trend and the periodicity of the fluctuation can distinguish the cooperative drift characteristics of trajectory deviation and height deviation in a gradual change state from the non-cooperative jump characteristics in an abrupt change state. This provides an accurate basis for subsequent differentiated correction based on operating condition identification, and avoids the problem of misaligned correction strategies caused by state misjudgment in existing technologies.

[0022] In one embodiment, the method first receives a structured light visual image of the weld area in front of the welding torch, and an arc signal from the end of the welding torch, the arc signal including transient values ​​of welding current and welding voltage. The structured light visual image is acquired by a structured light visual sensor installed at a predetermined distance in front of the welding torch. This sensor projects structured light stripes of a known pattern onto the weld surface and captures an image of deformed stripes modulated by the weld surface.

[0023] Here's a brief explanation of sensor selection. The structured light vision sensor can be a line laser scanner or a structured light camera with an area array. Its image output interface can use the GigE Vision protocol or USB 3.0 protocol. The image resolution is generally no less than 1280 pixels × 1024 pixels, and the frame rate is no less than 30 frames per second to ensure real-time extraction of weld features. The arc signal is read in real time by the feedback interface of the welding power supply. This feedback interface can be an analog output interface (voltage range 0 volts to 10 volts, corresponding to welding current 0 amperes to the rated maximum current) or a digital communication interface (such as RS-485 or CAN bus). The sampling frequency is synchronized with the frame rate of the structured light vision image, generally 500 to 1000 times per second, so that the two signals can correspond one-to-one on the time axis. Structured light vision has high accuracy in horizontal trajectory recognition, while arc sensing has fast response in the vertical direction. With the two complementing each other, even if a single sensor fails, the correction capability will not be completely lost, and the robustness of the system in complex welding environments is significantly improved.

[0024] The current three-dimensional trajectory coordinates and current welding torch height coordinates in the welding torch base coordinate system are obtained from the structured light visual image and the electric arc signal, respectively.

[0025] Specifically, the pixel coordinates of weld feature points are extracted from the structured light vision image and converted into the current three-dimensional trajectory coordinates in the welding torch base coordinate system according to the camera calibration parameters. The camera calibration parameters include intrinsic matrix (focal length, principal point coordinates, distortion coefficients) and extrinsic matrix (rotation matrix and translation vector of the camera relative to the welding torch base coordinate system). These parameters are pre-calibrated using a standard calibration plate method before the welding operation begins. The calibration process is a well-known technique and will not be elaborated here. The extraction of weld feature points employs a sub-pixel localization algorithm based on the stripe centerline. The centerline of the laser stripes is detected in the structured light image, and the weld center position is identified through the stripe deformation pattern, obtaining the pixel coordinates of the weld feature points in the image coordinate system. These coordinates are then converted to the welding torch base coordinate system using the calibration parameters, resulting in the current three-dimensional trajectory coordinates. The sub-pixel localization algorithm can employ the grayscale centroid method or Gaussian curve fitting method, which are image processing techniques well-known to those skilled in the art and will not be elaborated here. Structured light vision acquires three-dimensional trajectory coordinates, achieving sub-millimeter-level accuracy in horizontal weld tracking, thus meeting the stringent requirements of precision welding for trajectory localization.

[0026] Calculating the transient value of the arc length from the arc electrical signal includes: calculating the transient value of the arc resistance based on the ratio of the transient value of the welding voltage to the transient value of the welding current. In one embodiment, the transient value of the arc resistance is equal to the transient value of the welding voltage divided by the transient value of the welding current, and this calculation is performed in real time within each sampling period. Under preset welding process parameters, the transient value of the arc length is obtained by looking up the transient value of the arc resistance from a preset mapping table of arc resistance and arc length.

[0027] The mapping table was established through prior welding process experiments. Under constant shielding gas flow rate, wire extension length, and welding speed, there is a monotonic correspondence between arc resistance and arc length.

[0028] When establishing this mapping table, the true reference value of the arc length is obtained in the following way: Under the same welding process parameters, a laser displacement sensor (sampling frequency not less than 1kHz, measurement accuracy ±0.01mm) is used to measure the vertical distance from the end of the welding torch to the surface of the workpiece in real time, and the corresponding arc voltage and arc current values ​​are recorded at the same time. After repeated experiments, the average value is taken as the true reference value of the arc length corresponding to the arc resistance value.

[0029] The selection principle for data points is as follows: Within the arc resistance variation range under normal welding conditions (usually 0.01Ω to 0.10Ω, the specific value varies depending on the welding wire grade and shielding gas composition), 10 to 20 data points are selected at equal intervals according to the arc resistance value; if the arc length change rate (i.e., the absolute value of ΔL / ΔR) in a certain resistance range exceeds twice the average change rate of adjacent ranges, then the data points are densified in that range. After densification, the resistance interval between adjacent data points does not exceed 1 / 2 of the original equal interval to ensure that the nonlinear region of the mapping table has sufficient accuracy.

[0030] The verification method for the mapping table is as follows: After the initial calibration is completed, five sets of welding process parameter combinations that were not calibrated are randomly selected (within the above range of shielding gas flow rate, wire extension length, and welding speed). The arc length values ​​obtained by the table lookup method are compared with the measured values ​​of the laser displacement sensor. If the relative error of all verification points is within ±5%, the mapping table is considered qualified. If the relative error of any verification point exceeds ±5%, data points need to be added in the error excess area and the verification is repeated until all are qualified.

[0031] For example, the shielding gas flow rate can be 15~25L / min (preferably 20L / min), the wire extension length can be 15~25mm (preferably 20mm), and the welding speed can be 300~600mm / min (preferably 400mm / min). Those skilled in the art can select specific values ​​within the above range and recalibrate the mapping table according to the actual welding wire grade, shielding gas composition, and workpiece plate thickness.

[0032] The mapping table is stored in the system memory in the form of discrete data points. The number of data points is generally ten to twenty, covering the range of arc resistance variation under normal welding conditions. When actually looking up the table, linear interpolation is used to obtain the arc length value between the data points. The interpolation error is generally controlled within ±5%.

[0033] Based on the mapping relationship between the transient value of the arc length and the welding torch height, the current welding torch height coordinates are converted into the welding torch base coordinate system. The welding torch height coordinates are equal to the original installation height of the welding torch minus the transient value of the arc length, where the original installation height of the welding torch is the vertical distance between the tip of the welding torch and the workpiece surface at the start of welding, and this value is set during system initialization. By indirectly obtaining the welding torch height coordinates using the arc electrical signal, no additional height sensor needs to be installed at the tip of the welding torch, simplifying the structure; at the same time, the response speed of the arc signal to height changes is much faster than that of visual sensing, the correction delay in the height direction is controlled at the millisecond level, and fluctuations in the molten pool height can be quickly suppressed.

[0034] In one embodiment, the trajectory deviation vector between the current three-dimensional trajectory coordinates and the three-dimensional trajectory coordinates of the previous cycle, and the height deviation vector between the current welding torch height coordinates and the welding torch height coordinates of the previous cycle are calculated. The trajectory deviation vector is calculated by subtracting the three-dimensional trajectory coordinates of the previous cycle from the current three-dimensional trajectory coordinates. This vector describes the offset direction and offset distance of the welding torch tip in the horizontal plane relative to the position of the previous cycle. The height deviation vector is calculated by subtracting the welding torch height coordinates of the previous cycle from the current welding torch height coordinates. This vector describes the change in position of the welding torch tip in the vertical direction. The calculation of the above two deviation vectors is performed synchronously in each control cycle, which is generally four to ten milliseconds, to meet the real-time requirements. By calculating the deviation vectors of adjacent cycles instead of absolute coordinate deviations, the correction problem is transformed into a trend analysis problem. Subsequent state determination is based on dynamic characteristics rather than static thresholds, which can adapt to changes in working conditions at different welding speeds. The problem of fixed thresholds being too sensitive at low speeds and having a lag in response at high speeds is also avoided.

[0035] In one embodiment, it is determined whether the trajectory deviation vector and the height deviation vector have the same trend in time, and whether the fluctuation of the trajectory deviation vector is periodic or abrupt. When the trends are in the same direction and the fluctuation is periodic, the current weld is determined to be in a gradual change state; when the trends are not in the same direction or the fluctuation is abrupt, the current weld is determined to be in an abrupt change state. Specifically, the first change direction of the trajectory deviation vector within three consecutive cycles and the second change direction of the height deviation vector within the corresponding three cycles are calculated. The "three consecutive cycles" refer to three consecutive control cycles, with a corresponding time span of twelve to thirty milliseconds (calculated based on a control cycle of four to ten milliseconds). This time window is sufficient to cover the response time of the gradual change characteristics of the weld geometry, and is shorter than the duration of most abrupt events, effectively distinguishing between gradual change and abrupt change states. The method for determining the first change direction is: if the amplitude of the trajectory deviation vector in the current cycle is greater than the amplitude of the trajectory deviation vector in the previous cycle, the first change direction is determined to be increasing; if it is less, it is determined to be decreasing; if they are equal, the determination result of the previous cycle continues. The method for determining the second direction of change is the same as that for the first direction of change. When both the first and second directions of change are increasing or both are decreasing, the trends of change are determined to be in the same direction; if one increases while the other decreases, or one decreases while the other increases, the trends of change are determined to be in different directions.

[0036] The physical meaning of the criterion of unidirectional change trends is that when trajectory deviation and height deviation increase or decrease simultaneously, it indicates that the relative positions of the welding torch and the weld have changed in a coordinated manner in both the horizontal and vertical directions. This coordination usually stems from a slow, gradual change in weld geometry (such as the bevel angle gradually opening or closing), rather than a sudden jump. If the two trends are not in the same direction, it means that the deviation in at least one direction is caused by abrupt factors (such as molten pool spatter or sensor noise), and in this case, it should not be treated as a gradual change. Accumulated determination of the change direction over three consecutive cycles filters out accidental reversals caused by transient noise from the true coordinated trend, reducing the risk of misjudgment in single-cycle determinations and improving the reliability of state recognition.

[0037] The method for determining the periodicity of fluctuations is as follows: Calculate the number of fluctuation cycles of the trajectory deviation vector within the most recent ten cycles. The number of fluctuation cycles is calculated using the zero-crossing counting method, which counts the number of times the amplitude of the trajectory deviation vector changes from positive to negative or from negative to positive within ten cycles. Every two adjacent zero-crossings (one positive and one negative) are counted as one fluctuation cycle. When the number of fluctuation cycles is greater than or equal to three, the fluctuation is determined to be periodic. When the number of fluctuation cycles is less than three, if the rate of change of the amplitude of the trajectory deviation vector within a single control cycle exceeds a preset abrupt change threshold, the fluctuation is determined to be abrupt.

[0038] The threshold for the mutation is set as follows: Under normal welding speed (300 mm to 600 mm per minute), the rate of change of trajectory deviation caused by geometric mutations in the weld (such as right-angle corners, misalignment, and gap mutations) is generally greater than 20% to 30% per cycle. To verify the rationality of this threshold, the following comparative experiment was conducted: Under the same welding process parameters (welding current 280 amperes, welding voltage 28 volts, welding speed 400 mm per minute, shielding gas 80% argon plus 20% carbon dioxide, gas flow rate 20 liters per minute), three typical welds were tested respectively: (1) V-groove straight weld, groove angle 60 degrees, plate thickness 8 mm; (2) right-angle corner weld, corner angle 90 degrees, plate thickness 8 mm; (3) misaligned weld, misalignment 2 mm, plate thickness 8 mm. The test results show that the rate of change of trajectory deviation of the straight weld is stable at 5% to 10% per cycle, the rate of change at the right-angle corner is 35% to 45% per cycle, and the rate of change at the misalignment is 25% to 35% per cycle. Based on the experimental data above, the mutation threshold was set at 25% per cycle. This value falls within the transition range between gradual change and abrupt change, effectively distinguishing between the two states with a false positive rate of less than 5%. The above experiments were conducted under typical medium-thick plate welding conditions. For thin or thick plate welding, the mutation threshold can be adjusted within the range of 20% to 30% according to the actual process parameters.

[0039] If the fluctuation does not exceed the aforementioned mutation threshold, it is determined that the fluctuation exhibits a non-periodic disturbance and is treated as a mutation state. A non-periodic disturbance refers to small, irregular deviation fluctuations caused by random noise or sensor jitter, characterized by a rate of change lower than the mutation threshold and no obvious periodicity. Since such disturbances lack predictable, gradual changes, treating them as gradual changes would lead to an overreaction of the correction system to noise. Therefore, they are treated as mutation states, maintaining the trajectory baseline unchanged and correcting only the height to avoid introducing unnecessary horizontal oscillations. By combining the fluctuation period number and the rate of change threshold, three fundamentally different fluctuation modes—predictable periodicity of gradual changes, drastic jumps in mutation states, and randomness of non-periodic disturbances—are distinguished. The correction strategy precisely matches the actual geometric characteristics of the weld, avoiding the oscillation and jitter problems caused by the "one-size-fits-all" correction strategy in existing technologies.

[0040] When the condition is determined to be a gradual change, the slow alteration of weld geometry causes the trajectory deviation and height deviation to exhibit a coordinated drift characteristic. In this case, the essence of the trajectory deviation is not that the welding torch has deviated from its correct position, but rather that the weld centerline itself has undergone an objective drift. Forcibly correcting the trajectory deviation will cause the correction system to counteract the thermal deformation trend, resulting in continuous horizontal oscillations.

[0041] Therefore, when in a state of gradual change, a first comprehensive trajectory correction amount is generated from the current three-dimensional trajectory coordinates and the current welding torch height coordinates. The first comprehensive trajectory correction amount only corrects the height deviation vector, while the trajectory deviation vector remains unchanged at its current value.

[0042] Specifically, a height correction component is generated from the current welding torch height coordinates and the height deviation vector; the trajectory deviation vector does not participate in the generation of the correction component; the height correction component is used as the first comprehensive trajectory correction amount. Under gradual change conditions, the geometric changes of the weld are mainly manifested as slow changes in the bevel angle or assembly gap. These changes primarily affect the welding torch height (i.e., the vertical distance between the welding torch and the workpiece surface), with little impact on the horizontal trajectory position. By correcting only the height deviation, the horizontal oscillations introduced by simultaneously correcting the trajectory deviation are avoided, thus maintaining the stability of the welding process.

[0043] The phrase "keeping the trajectory deviation vector unchanged" means that the trajectory deviation vector is used as the position drift reference of the welding torch tip relative to the weld centerline. Specifically, the trajectory deviation vector is decomposed into a direction component and an amplitude component in the horizontal plane. The direction component serves as the position drift direction of the welding torch tip in the horizontal plane, and the amplitude component serves as the position drift distance of the welding torch tip in the horizontal plane. The direction component maintains a constant cumulative value during the duration of the gradual change state. This cumulative value specifically refers to the initial value of the direction component locked from the first control cycle of entering the gradual change state, and it is not updated with local fluctuations in the weld in subsequent cycles until the state ends or a reset condition is triggered, i.e., it is not recalculated with each cycle update. The amplitude component is updated according to the amplitude of the current trajectory deviation vector during the duration of the gradual change state. That is, in each control cycle, the amplitude of the current trajectory deviation vector is used as the new amplitude component value, thereby reflecting the latest change in the drift distance of the weld centerline. However, when the gradual change state continues to exceed the preset reference holding period, or when the change trend is detected to be reversed, the position drift reference is reset as a whole to the current trajectory deviation vector, that is, the direction component and the amplitude component are reassigned with the corresponding components of the current trajectory deviation vector.

[0044] The reference holding period number is set based on the following: the reference holding period number determines the length of time the position drift reference is maintained under a gradual change. If the holding time is too short, the welding torch frequently resets the reference, and the cumulative effect of thermal deformation cannot be effectively tracked; if the holding time is too long, the reference fails to be updated in time when the actual weld direction changes significantly, resulting in tracking lag. To determine a reasonable reference holding period number, the following comparative experiment was conducted: Welding was performed on a low-carbon steel plate with a thickness of 8 mm and a V-groove angle of 60 degrees. The welding speed was 400 mm / min, the welding current was 280 amperes, and the welding voltage was 28 volts. The reference holding period number was set to 20, 50, 100, and 150 control cycles (the control cycle is 4 milliseconds, corresponding to actual times of 80 milliseconds, 200 milliseconds, 400 milliseconds, and 600 milliseconds, respectively), and the standard deviation of the welding torch trajectory tracking error was measured.

[0045] Experimental results show that the standard deviation of the tracking error is 0.8 mm for 20 cycles, 0.3 mm for 50 cycles, 0.2 mm for 100 cycles, and 0.35 mm for 150 cycles. Based on the above data, a reference hold period of 50 to 100 control cycles (corresponding to an actual time of 200 to 400 milliseconds) is the preferred range, as this range results in the smallest tracking error and the best stability.

[0046] The reference holding period can also be dynamically adjusted according to the welding speed. The faster the welding speed, the fewer the reference holding periods. The specific adjustment relationship is: Reference holding period N = (L / v) / T, where N is the reference holding period (dimensionless integer), L is the fixed distance (30mm~50mm, preferably 40mm), v is the current welding speed (the unit must be consistent in mm / ms when calculating, for example, 400mm / min can be converted to 6.67mm / ms), and T is the control period (4ms~10ms). The reference holding period is equal to the fixed distance (generally 30mm to 50mm) divided by the welding speed and then divided by the control period, so that the reference holding time remains consistent under different welding speeds. The fixed distance of 30mm to 50mm is selected based on the fact that within the range of welding speed of 300mm to 600mm per minute, the reference holding time corresponding to this distance is within the preferred range of 50 to 100 control periods. When the gradual change continues for more than a preset reference holding period, or when a reversal of the change trend is detected, the position drift reference is reset to the current trajectory deviation vector. The trajectory deviation vector is decomposed into directional and amplitude components and processed separately. The cumulative holding of the directional component allows the welding torch to adapt to changes in the overall weld orientation, while the periodic updating of the amplitude component allows the welding torch to track the latest changes in the weld centerline drift distance. The combination of these two methods avoids frequent reference adjustments due to local noise and ensures stable tracking during the thermal deformation accumulation process.

[0047] When a sudden change is detected, the weld geometry changes drastically in the horizontal direction, requiring a rapid response to trajectory deviations to prevent the welding torch from deviating from the weld centerline. However, at this time, the molten pool surface oscillates violently due to the sudden change in heat input, the arc length fluctuates wildly, and the height deviation is mixed with a large amount of noise. Simultaneously correcting the height causes the welding torch to shake up and down following the molten pool surface, introducing instability in the vertical direction.

[0048] Therefore, when in a sudden change state, a second comprehensive trajectory correction amount is generated from the current three-dimensional trajectory coordinates and the current welding torch height coordinates. This second comprehensive trajectory correction amount only corrects the trajectory deviation vector, while the height deviation vector remains unchanged. Specifically, a trajectory correction component is generated from the current three-dimensional trajectory coordinates and the trajectory deviation vector; the height deviation vector does not participate in the generation of the correction component; and the trajectory correction component is used as the second comprehensive trajectory correction amount.

[0049] In abrupt changes, the weld geometry undergoes rapid transformations (such as cornering, misalignment, and gap abrupt changes). Horizontal trajectory deviation becomes the primary concern, requiring a rapid response to prevent the welding torch from deviating from the weld centerline. Height deviations at the moment of abrupt change are often accompanied by significant noise (such as arc length fluctuations caused by molten pool spatter). Simultaneous height correction might introduce vertical jitter; therefore, the height deviation is kept constant until the abrupt change ends, at which point height correction is resumed. The height deviation vector remaining constant means that its value is locked at the moment of entering the abrupt change state, no longer updated periodically, and not included in the calculation of correction components. Instead, it is directly output to subsequent cycles as the height oscillation reference for the welding torch tip relative to the current molten pool surface. Subsequent height compensation uses this reference as its origin, rather than forcibly correcting the height deviation to zero. By using the height deviation as an oscillation reference and resetting it during state transitions, a smooth transition between height stability during abrupt changes and height tracking during gradual changes is achieved. Height jumps during state transitions are avoided, ensuring the continuity of the welding process under different operating conditions.

[0050] In one embodiment, the process of generating a height correction component from the current welding torch height coordinates and the height deviation vector further includes weight calculation. Specifically, the height correction weight is calculated as the ratio of the amplitude of the height deviation vector to the amplitude of the trajectory deviation vector. When the amplitude of the trajectory deviation vector is zero, the height correction weight is set to a preset maximum weight value, typically set to 0, indicating that the height correction is unaffected by the trajectory deviation. When the amplitude of the trajectory deviation vector is not zero, the height correction weight is equal to the amplitude of the height deviation vector divided by the amplitude of the trajectory deviation vector.

[0051] The weight threshold is set based on the following criteria: the weight threshold determines whether the height correction component is generated directly or after smoothing. To determine a reasonable weight threshold, the following comparative experiment was conducted: welding was performed on a low-carbon steel plate with a thickness of 8 mm and a V-groove angle of 60 degrees at a welding speed of 400 mm per minute. The weight thresholds were set to 0.3, 0.5, 0.7, and 0.9, respectively, and the response time and steady-state error of the height correction were measured.

[0052] Experimental results show that with a weight threshold of 0.3, the response time is 8 milliseconds but the steady-state error is 0.15 millimeters; with a weight threshold of 0.5, the response time is 12 milliseconds and the steady-state error is 0.05 millimeters; with a weight threshold of 0.7, the response time is 18 milliseconds and the steady-state error is 0.03 millimeters; and with a weight threshold of 0.9, the response time is 25 milliseconds and the steady-state error is 0.02 millimeters. Considering both response speed and steady-state accuracy, a weight threshold of 0.5 is the preferred value, achieving the best balance between response time and steady-state error. When the height correction weight is greater than the preset weight threshold, the height correction component is directly generated from the current welding torch height coordinates and the height deviation vector, i.e., the height correction component equals the height deviation vector multiplied by the height correction weight. When the height correction weight is less than or equal to the weight threshold, the height correction component is generated from the current welding torch height coordinates and the height deviation vector after smoothing.

[0053] The smoothing process employs a first-order low-pass filter algorithm, with filter coefficients typically ranging from 0.3 to 0.7. The selection of the filter coefficients is based on the following: the filter coefficient determines the strength of the smoothing process; a smaller coefficient results in a stronger smoothing effect but also a greater response delay. To determine a reasonable range for the filter coefficients, the following comparative experiments were conducted: Under the same welding conditions, filter coefficients of 0.1, 0.3, 0.5, 0.7, and 0.9 were set, and the signal-to-noise ratio (SNR) and phase delay of the height-corrected signal were measured. Experimental results show that with a filter coefficient of 0.1, the SNR is 12 dB but the phase delay is 15 ms; with 0.3, the SNR is 18 dB and the phase delay is 8 ms; with 0.5, the SNR is 22 dB and the phase delay is 5 ms; with 0.7, the SNR is 20 dB and the phase delay is 3 ms; and with 0.9, the SNR is 15 dB and the phase delay is 1 ms. Considering both signal-to-noise ratio (SNR) and phase delay, a filter coefficient range of 0.3 to 0.7 is preferred. Within this range, the SNR is higher than 18 dB and the phase delay is lower than 8 milliseconds, meeting the requirements for real-time correction. When the altitude deviation is small relative to the trajectory deviation, it indicates that altitude change is a secondary factor, and smoothing is used to suppress noise. When altitude deviation is dominant, direct correction is performed to ensure response speed. The altitude correction weight is adaptively adjusted based on its relative magnitude with the trajectory deviation, enabling dynamic optimization of the altitude correction strength. This results in a fast response when altitude changes are significant and noise suppression when altitude changes are minor, avoiding over-correction or under-correction problems caused by fixed correction gain.

[0054] In one embodiment, the process of generating a trajectory correction component from the current 3D trajectory coordinates and the trajectory deviation vector also includes weight calculation. Specifically, the trajectory correction weight is calculated as the ratio of the magnitude of the trajectory deviation vector to the magnitude of the height deviation vector. When the magnitude of the height deviation vector is zero, the trajectory correction weight is set to a preset maximum weight value, typically set to 0. When the magnitude of the height deviation vector is not zero, the trajectory correction weight is equal to the magnitude of the trajectory deviation vector divided by the magnitude of the height deviation vector. When the trajectory correction weight is greater than a preset weight threshold, the trajectory correction component is directly generated from the current 3D trajectory coordinates and the trajectory deviation vector, i.e., the trajectory correction component is equal to the trajectory deviation vector multiplied by the trajectory correction weight. When the trajectory correction weight is less than or equal to the weight threshold, the trajectory correction component is generated from the current 3D trajectory coordinates and the trajectory deviation vector after amplitude limiting processing.

[0055] The maximum correction amplitude is set based on the following: the maximum correction amplitude limits the extreme displacement of trajectory correction under abrupt changes, preventing over-correction due to sensor malfunction or calculation error. To determine a reasonable maximum correction amplitude, the following comparative experiment was conducted: welding was performed on a right-angle corner weld with a plate thickness of 8 mm at a welding speed of 400 mm / min. The maximum correction amplitude was set to 0.3 mm, 0.5 mm, 0.8 mm, 1.0 mm, and 1.5 mm, respectively, and the overshoot and settling time of the welding torch trajectory were measured. The experimental results show that at a maximum correction amplitude of 0.3 mm, the overshoot is 0.1 mm but the settling time is 40 milliseconds; at 0.5 mm, the overshoot is 0.2 mm and the settling time is 25 milliseconds; at 0.8 mm, the overshoot is 0.4 mm and the settling time is 15 milliseconds; at 1.0 mm, the overshoot is 0.6 mm and the settling time is 10 milliseconds; and at 1.5 mm, the overshoot is 1.2 mm and the settling time is 8 milliseconds, but oscillations occur. Considering both overshoot and settling time, the preferred maximum correction amplitude is between 0.5 mm and 1.0 mm, within which tracking response is rapid and without significant oscillations. The amplitude limiting method is as follows: if the amplitude of the trajectory correction component exceeds the preset maximum correction amplitude, it is limited to the maximum correction amplitude range. Under abrupt changes, overcorrection due to sensor noise or calculation errors is prevented, and violent welding torch oscillations are avoided. The trajectory correction weight is adaptively adjusted based on its relative magnitude with the height deviation, achieving dynamic optimization of the trajectory correction intensity. When the trajectory deviation is significant, a rapid response is achieved to avoid deviation from the weld seam; when the trajectory deviation is weak, the correction amplitude is limited to prevent oscillations. Tracking accuracy is ensured, and the smoothness of the correction process is also guaranteed.

[0056] In one embodiment, the first or second comprehensive trajectory correction value is fed forward to the motion controller of the welding robot to drive the welding torch tip to perform six-degree-of-freedom trajectory compensation. The first or second comprehensive trajectory correction value is output in the form of a three-dimensional spatial vector. This vector describes the desired displacement of the welding torch tip in the X, Y, and Z directions, as well as the desired rotation angles (i.e., pitch, yaw, and roll angles) around the X, Y, and Z axes in the welding torch base coordinate system. After receiving the correction value, the motion controller decomposes it into incremental angle commands for each joint axis through inverse kinematics calculation, and sends them to each joint motor via servo drivers. The six-degree-of-freedom compensation of the welding torch tip is thus realized in real time. The communication interface between the motion controller and the correction value generation unit can adopt the EtherCAT protocol or the Modbus TCP protocol. The communication cycle is synchronized with the control cycle (four to ten milliseconds). The correction value is transmitted in a 16-bit signed integer or 32-bit floating-point format. Each component corresponds to the displacement in the X, Y, and Z directions (in micrometers or millimeters) and the rotation angle around each axis (in milliradians or degrees). Feedforward control means that the correction is directly added to the robot's original trajectory planning instructions, rather than waiting for deviations to occur through a feedback loop before responding. The correction delay is controlled within a single control cycle, which meets the stringent response speed requirements of real-time welding. The correction is directly injected into the motion controller, reducing the correction delay from tens of milliseconds in traditional feedback control to a single control cycle (four to ten milliseconds). The welding torch begins to respond the instantaneously when a geometric change occurs in the weld seam, thus avoiding tracking overshoot and correction overshoot problems caused by feedback lag.

[0057] In one embodiment, optionally, the method further includes a quality assessment step for the structured light visual image. After receiving the structured light visual image, the image's contrast and stripe sharpness indices are calculated. If the contrast is below a preset threshold or stripes are broken, it is determined that the visual sensor is contaminated or subject to illumination interference. In this case, the weight of the arc electrical signal in the trajectory coordinate calculation is automatically increased, while the weight of the structured light visual image is decreased to ensure the continuity of the correction. The contrast threshold is generally taken as 30% to 50% of the image's grayscale dynamic range. For example, for an eight-bit grayscale image (grayscale range 0 to 255), the contrast threshold can be taken as 80% to 120%. After the visual image quality returns to normal, the original weight allocation is gradually restored. With visual quality assessment and adaptive weight switching, the basic correction capability can still be maintained under extreme conditions of structured light visual failure, and the system's usability in harsh welding environments is significantly improved.

[0058] In one embodiment, optionally, the method further includes a filtering preprocessing step for the arc electrical signal. Before calculating the transient value of the arc resistance, the transient values ​​of the welding current and welding voltage are respectively subjected to moving average filtering. The filtering window length is generally five to ten sampling periods to eliminate the interference of high-frequency arc noise on the resistance calculation. The filtered current and voltage values ​​are then used to calculate the transient value of the arc resistance, thereby improving the stability of the arc length estimation. The moving average filtering preprocessing reduces the fluctuation in resistance calculation caused by arc noise by about 60% to 80%, significantly improves the stability of the arc length estimation, and reduces highly corrected false triggers.

[0059] In one embodiment, optionally, the method further includes a hysteresis compensation step for the weld condition determination result. A structured light vision sensor is mounted in front of the welding torch, and the weld condition detected by the sensor has a spatial lead-in relative to the current position of the welding torch. This spatial lead-in is typically between 50 and 150 millimeters, with the specific value depending on the installation position of the structured light vision sensor and can be determined by those skilled in the art based on the actual welding torch structure. After determining the weld condition, the time delay for the condition to reach the welding torch position is calculated based on the welding speed and the lead-in distance. This delayed condition determination result is used to generate a correction amount to avoid timing deviations in corrections caused by the state determination being ahead of the weld at geometric abrupt changes. Hysteresis compensation ensures that the state determination and the actual position of the welding torch are aligned in time, avoiding premature corrections caused by visual lead-in. The welding torch only executes the correction strategy for abrupt changes when it reaches a corner, preventing weld deviations caused by misaligned correction timing.

[0060] In one embodiment, optionally, the method further includes an output limiting protection step for the correction amount. Before outputting the first or second integrated trajectory correction amount to the motion controller, the amplitude and rate of change of each component of the correction amount are checked separately. If the amplitude of a component exceeds the maximum permissible correction amplitude for a single axis, or if the rate of change of a component in an adjacent period exceeds the maximum permissible rate of change for a single axis, the component is limited to the permissible range, and an abnormal alarm signal is triggered, prompting the operator to check the sensor status or welding process parameters. Output limiting protection ensures safe operation even under sensor failure or extreme conditions, and prevents welding torch collisions or weld overheating accidents caused by abnormal correction amounts.

[0061] Through the close collaboration of the above steps, this invention achieves deep integration of structured light vision and arc electrical signals during the welding process. Furthermore, by intelligently determining the weld condition and employing differentiated correction strategies, it overcomes the limitations of single sensors and the simplistic correction strategies in existing technologies. In a gradually changing state, by maintaining a trajectory deviation benchmark and focusing on height correction, over-tracking in the horizontal direction is avoided. In abrupt changes, by maintaining a height benchmark and focusing on trajectory correction, rapid response to geometrical changes in the weld is achieved. This state-specific, refined correction mechanism allows the welding robot to balance tracking accuracy and process stability when facing complex and varied weld conditions, thereby improving welding quality.

[0062] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for real-time correction of welding trajectories based on integrated vision guidance, characterized in that, include: Receive structured light visual images of the weld area in front of the welding torch, as well as electric arc signals from the end of the welding torch, wherein the electric arc signals include transient values ​​of welding current and transient values ​​of welding voltage; The current three-dimensional trajectory coordinates and current welding torch height coordinates in the welding torch base coordinate system are obtained from the structured light visual image and the arc electrical signal, respectively. Calculate the trajectory deviation vector between the current three-dimensional trajectory coordinates and the three-dimensional trajectory coordinates of the previous cycle, and the height deviation vector between the current welding torch height coordinates and the welding torch height coordinates of the previous cycle; It is determined whether the trajectory deviation vector and the height deviation vector have the same trend in time, and whether the fluctuation of the trajectory deviation vector is periodic or abrupt; the abruptness refers to the rate of change of the amplitude of the trajectory deviation vector within a single control cycle exceeding a preset abruptness threshold; when the trends are in the same direction and the fluctuation is periodic, the current weld is determined to be in a gradual change state; when the trends are not in the same direction or the fluctuation is abrupt, the current weld is determined to be in an abrupt state. When in a gradual change state, a first comprehensive trajectory correction amount is generated from the current three-dimensional trajectory coordinates and the current welding torch height coordinates. The first comprehensive trajectory correction amount only corrects the height deviation vector, and the trajectory deviation vector remains unchanged. When in a sudden change state, a second comprehensive trajectory correction amount is generated from the current three-dimensional trajectory coordinates and the current welding torch height coordinates. The second comprehensive trajectory correction amount only corrects the trajectory deviation vector, and the height deviation vector remains unchanged. The first comprehensive trajectory correction amount or the second comprehensive trajectory correction amount is fed forward to the motion controller of the welding robot to drive the end of the welding torch to perform six-degree-of-freedom trajectory compensation.

2. The method for real-time correction of welding trajectory based on integrated vision guidance according to claim 1, characterized in that, The step of determining whether the trajectory deviation vector and the altitude deviation vector exhibit the same trend in time, and determining whether the fluctuation of the trajectory deviation vector shows periodicity or abruptness, includes: Calculate the first direction of change of the trajectory deviation vector within three consecutive cycles, and the second direction of change of the height deviation vector within the corresponding three cycles; when both the first direction of change and the second direction of change are increasing or both are decreasing, it is determined that the changing trends are in the same direction. Calculate the number of fluctuation cycles of the trajectory deviation vector in the most recent ten cycles; when the number of fluctuation cycles is greater than or equal to three, the fluctuation is determined to be periodic; when the number of fluctuation cycles is less than three, if the rate of change of the amplitude of the trajectory deviation vector in a single control cycle exceeds a preset mutation threshold, the fluctuation is determined to be abrupt; if it does not exceed the mutation threshold, the fluctuation is determined to be a non-periodic disturbance and is processed according to the abrupt state.

3. The method for real-time correction of welding trajectory based on integrated vision guidance according to claim 1, characterized in that, When in a state of gradual change, the generation of a first comprehensive trajectory correction amount from the current three-dimensional trajectory coordinates and the current welding torch height coordinates includes: A height correction component is generated from the current welding torch height coordinates and the height deviation vector; The trajectory deviation vector does not participate in the generation of the correction component; The height correction component is used as the first comprehensive trajectory correction amount.

4. The method for real-time correction of welding trajectory based on integrated vision guidance according to claim 3, characterized in that, When in a sudden change state, the generation of a second comprehensive trajectory correction amount from the current three-dimensional trajectory coordinates and the current welding torch height coordinates includes: A trajectory correction component is generated from the current three-dimensional trajectory coordinates and the trajectory deviation vector; The height deviation vector does not participate in the generation of the correction component; The trajectory correction component is used as the second comprehensive trajectory correction amount.

5. The method for real-time correction of welding trajectory based on integrated vision guidance according to claim 1, characterized in that, The step of obtaining the current three-dimensional trajectory coordinates and current welding torch height coordinates in the welding torch base coordinate system from the structured light visual image and the arc electrical signal respectively includes: The pixel coordinates of weld feature points are extracted from the structured light visual image and converted into the current three-dimensional trajectory coordinates in the welding torch base coordinate system according to the camera calibration parameters; Calculating the transient value of arc length from the arc electrical signal includes: calculating the transient value of arc resistance based on the ratio of the transient value of welding voltage to the transient value of welding current; under preset welding process parameters, obtaining the transient value of arc length from the transient value of arc resistance by looking up the table based on the preset mapping relationship table between arc resistance and arc length; and converting the transient value of arc length into the current welding torch height coordinates in the welding torch base coordinate system based on the mapping relationship between the transient value of arc length and welding torch height.

6. The method for real-time correction of welding trajectory based on integrated vision guidance according to claim 1, characterized in that, When in a state of gradual change, the trajectory deviation vector remains unchanged at its current value, including: The trajectory deviation vector is used as the position drift reference of the welding torch tip relative to the weld centerline. The trajectory compensation of the welding torch tip in subsequent cycles is based on the position drift reference as the origin. When the gradual change state continues for more than the preset reference holding period, or when the change trend is detected to be reversed, the position drift reference is reset to the current trajectory deviation vector.

7. The method for real-time correction of welding trajectory based on integrated vision guidance according to claim 1, characterized in that, The height deviation vector remains unchanged, including: The height deviation vector is used as the height oscillation reference of the welding torch tip relative to the current molten pool surface. The height compensation of the welding torch tip in subsequent cycles is based on the height oscillation reference as the origin. When the abrupt change state ends and the gradual change state begins, the height oscillation reference is reset to the current height deviation vector.

8. The method for real-time correction of welding trajectory based on integrated vision guidance according to claim 3, characterized in that, The generation of a height correction component from the current welding torch height coordinates and the height deviation vector includes: The height correction weight is calculated from the ratio of the height deviation vector to the trajectory deviation vector; When the magnitude of the trajectory deviation vector is zero, the height correction weight is set to the preset maximum weight value; When the magnitude of the trajectory deviation vector is not zero, the height correction weight is equal to the magnitude of the height deviation vector divided by the magnitude of the trajectory deviation vector. When the height correction weight is greater than the preset weight threshold, the height correction component is directly generated from the current welding torch height coordinates and the height deviation vector; When the height correction weight is less than or equal to the weight threshold, the height correction component is generated by smoothing the current welding torch height coordinates and the height deviation vector.

9. A real-time correction method for welding trajectory based on integrated vision guidance according to claim 4, characterized in that, The generation of trajectory correction components from the current three-dimensional trajectory coordinates and the trajectory deviation vector includes: The trajectory correction weight is calculated based on the ratio of the trajectory deviation vector to the height deviation vector; When the magnitude of the height deviation vector is zero, the trajectory correction weight is set to the preset maximum weight value; When the magnitude of the height deviation vector is not zero, the trajectory correction weight is equal to the magnitude of the trajectory deviation vector divided by the magnitude of the height deviation vector. When the trajectory correction weight is greater than a preset weight threshold, the trajectory correction component is directly generated from the current three-dimensional trajectory coordinates and the trajectory deviation vector; When the trajectory correction weight is less than or equal to the weight threshold, the trajectory correction component is generated by the current three-dimensional trajectory coordinates and the trajectory deviation vector after amplitude limiting processing. The amplitude limiting processing refers to limiting the amplitude of the trajectory correction component to a preset maximum correction amplitude range, where the maximum correction amplitude is 0.5 mm to 1.0 mm.

10. A real-time welding trajectory correction method based on integrated vision guidance according to claim 6, characterized in that, Using the trajectory deviation vector as a position drift reference for the welding torch tip relative to the weld centerline includes: The trajectory deviation vector is decomposed into a direction component and an amplitude component in the horizontal plane; The directional component serves as the positional drift direction of the welding torch tip in the horizontal plane. The amplitude component is the positional drift distance of the welding torch tip in the horizontal plane; The directional component retains its cumulative value during the duration of the gradual change state; The amplitude component is updated according to the amplitude of the current trajectory deviation vector.

Images