Welding gun posture self-adaptive control method and system based on super-thick wall pipeline welding

Abstract

The application discloses a welding gun posture self-adaptive control method and system based on super-thick wall pipeline welding and relates to the technical field of pipeline welding.The welding gun posture self-adaptive control method and system based on super-thick wall pipeline welding comprises the following steps: S1, preprocessing original observation data and previous control record data; S2, identifying space constraint difference, center offset features and molten pool response offset features, performing state discrimination and compensation level output; S3, performing posture correction and heat input area regression control according to equivalent heat input center reconstruction analysis results; and S4, performing closed-loop return control and welding maintenance control according to convergence control analysis results.The method solves the problem that in the prior art, the geometric center of a groove or the nominal center of a weld is often used for welding gun centering and posture control in the super-thick wall pipeline welding process, the arc heat input center offset cannot be effectively constrained under the condition that the distance between the two sides of the groove changes asymmetrically, and the two sides are prone to uneven heating and fusion.

Description

Technical Field

[0001] This invention belongs to the field of pipeline welding technology, specifically a welding torch posture adaptive control method and system for ultra-thick-walled pipeline welding. Background Technology

[0002] Ultra-thick-walled pipelines are widely used in oil and gas transportation, chemical plants, nuclear power equipment, pressure vessels, and large-scale energy projects. Their welding quality directly affects the pressure-bearing capacity, sealing performance, and service safety of pipeline connections. Due to the characteristics of ultra-thick-walled pipelines, such as large wall thickness, deep bevels, numerous welding layers, long heat input duration, and complex spatial variations in welding positions, existing projects typically employ automated welding equipment, robotic welding equipment, or semi-automatic welding equipment to complete multi-layer, multi-pass welding operations. Existing welding control methods generally revolve around weld seam tracking, welding torch position control, welding torch posture adjustment, and coordination of welding process parameters. Welding process information is acquired through visual inspection devices, position detection devices, and welding electrical parameter monitoring devices. The forward tilt, backward drag, lateral movement, and trajectory of the welding torch are then adjusted to adapt to the welding requirements of different welding positions and different pass conditions. With the development of image processing, sensor detection, robot motion control and process monitoring technologies, the automation level of ultra-thick wall pipe welding process is constantly improving. The welding system's ability to perceive and adjust the groove shape, welding torch posture, arc state and molten pool shape online is also gradually being enhanced. Related technologies have become important technical means to improve the welding quality and operation efficiency of ultra-thick wall pipes.

[0003] For example, invention patent CN111571048B discloses a method, device, storage medium, and processor for determining pipeline welding speed. The method includes: conducting a predetermined number of welding tests on the pipeline to obtain test results; acquiring historical data based on the test results, including historical welding speed and historical attitude angle; acquiring the real-time attitude angle of the welding torch at various positions on the pipeline; and determining the real-time welding speed of the welding torch at each position based on the historical data and the real-time attitude angle. Specifically, a fitting curve is obtained by fitting the historical welding speed and historical attitude angle data, and the functional relationship between the real-time welding speed and the real-time attitude angle is determined based on the fitting curve. The real-time welding speed of the welding torch at each position is then determined based on this functional relationship and the real-time attitude angle. The method also discloses a corresponding determining device, storage medium, processor, and welding device, realizing dynamic determination of pipeline welding speed based on changes in the welding torch attitude angle, thus improving the adaptability of welding speed control during pipeline welding.

[0004] For example, the invention patent with publication number CN119387995B discloses an online planning method and related equipment for welding torch posture in multi-layer, multi-pass external welding of pipelines. The method includes: for each filler weld pass during pipeline external welding, acquiring laser line images of the pipeline welding bevel and welding torch posture data; extracting feature pixels from the laser line images; converting the feature pixels into a three-dimensional point cloud using a detection mathematical model and fitting it to obtain a three-dimensional welding bevel image; calculating welding bevel size parameters, welding torch posture parameters, and workpiece posture parameters using the three-dimensional welding bevel image and welding torch posture data; and determining the current welding bevel posture based on the welding bevel size parameters, welding torch posture parameters, and workpiece posture parameters. The method involves planning the welding torch posture for each pass, controlling the welding torch to perform filler welding along the welding bevel according to the planned posture, and detecting and correcting the welding torch posture parameters in real time during the filler welding process. During the return stroke after each filler welding pass, the method acquires the return laser line image of the welding bevel containing the weld bead and the return welding torch posture data. The method then uses the return laser line image and return welding torch posture data to plan the welding torch posture for the next pass. After completing multiple filler welding passes, a capping weld is performed to complete multi-layer, multi-pass external welding of the pipeline. The method also discloses corresponding devices, welding equipment, and electronic equipment, enabling online planning and real-time correction of the welding torch posture during multi-layer, multi-pass external welding of pipelines.

[0005] While existing pipeline welding control technologies can adjust the welding process by combining welding torch posture, welding speed, bevel image, and pose planning results, they still largely focus on controlling the welding torch's trajectory and posture based on bevel geometry, weld center position, or preset pose parameters. They pay more attention to welding speed matching, online welding torch pose planning, and trajectory tracking itself. However, they lack a dedicated constraint mechanism for the arc heat input center offset problem caused by asymmetrical changes in the bevel spacing, interlayer morphology evolution, and local spatial constraint differences during ultra-thick wall pipeline welding. This makes it difficult to synchronously coordinate the relationship between the welding torch alignment, arc position, and molten pool response position, which can easily lead to uneven heating on both sides, inconsistent fusion states, and weld bead skew in complex welding sections.

[0006] Therefore, in order to address the above problems, there is an urgent need for an adaptive control method and system for welding torch posture based on ultra-thick-walled pipe welding. Summary of the Invention

[0007] To address the above problems, this invention provides a welding torch posture adaptive control method and system for ultra-thick-walled pipe welding. This method solves the problem that in the existing technology for ultra-thick-walled pipe welding, the welding torch centering and posture control are mostly based on the geometric center of the groove or the nominal center of the weld. This makes it difficult to effectively constrain the offset of the arc heat input center under the condition of asymmetrical changes in the distance between the two sides of the groove, which easily leads to uneven heating and fusion on both sides.

[0008] To achieve the above objectives, the technical solution adopted by this invention is: an adaptive control method for welding torch posture based on ultra-thick-walled pipe welding, comprising: S1, collecting raw observation data during the operation of the welding equipment, acquiring preceding control record data, and preprocessing the raw observation data and preceding control record data; S2, jointly identifying the spatial constraint differences on both sides of the bevel, the arc center offset characteristics, and the molten pool response offset characteristics based on the raw observation data, and performing edge state discrimination and compensation level output; S3, performing equivalent heat input center reconstruction analysis on the raw observation data and preceding control record data, and performing welding torch posture correction and heat input area regression control based on the equivalent heat input center reconstruction analysis results; S4, executing control based on the welding torch posture correction results and performing convergence control analysis, and performing closed-loop feedback control and weld holding control based on the convergence control analysis results.

[0009] Furthermore, the specific process of collecting raw observation data during the operation of the welding equipment and obtaining the preceding control record data is as follows: During the operation of the welding equipment, the raw observation data corresponding to the current welding section is received and organized. The raw observation data includes: the coordinate data of the contour points of the left wall of the bevel, the coordinate data of the contour points of the right wall of the bevel, the coordinate data of the center line of the welding trajectory, the coordinate data of the center of the welding torch end, the forward tilt angle data of the welding torch, the backward drag angle data of the welding torch, the lateral swing angle data of the welding torch, the spatial coordinate data of the welding wire end, the arc spot image data, the molten pool contour image data, the welding current data, the welding voltage data, the wire feed speed data, the welding speed data, the circumferential position angle data of the welding torch, the current layer pass number data, and the surface contour point data of the previous layer pass. The previous layer refers to the welding layer above the welding layer corresponding to the current layer pass number data. The preceding control record data is then obtained and a preceding control record database is established.

[0010] Furthermore, the specific preprocessing steps for the raw observation data and preceding control record data are as follows: The raw observation data undergoes unified timestamp alignment; invalid frames are removed, missing frames are filled, and continuous frames are smoothed for the arc spot image data and molten pool contour image data; the surface contour point data of the previous weld bead and the preceding control record data undergo circumferential position matching and processing; the raw observation data and the preceding control record data are standardized using the maximum absolute value standardization algorithm; and the coordinate data of the left side wall contour point of the bevel, the right side wall contour point of the bevel, the center line coordinates of the welding trajectory, the center coordinates of the welding torch tip, the forward tilt angle of the welding torch, the backward drag angle of the welding torch, the lateral sway angle of the welding torch, the spatial coordinates of the welding wire tip, and the circumferential position angle of the welding torch are normalized using the maximum and minimum value normalization algorithm.

[0011] Furthermore, the specific process for jointly identifying the spatial constraint differences on both sides of the bevel, the arc center offset characteristics, and the molten pool response offset characteristics based on the original observation data is as follows: Based on the coordinate data of the contour points on the left and right sides of the bevel and the spatial coordinate data of the welding wire end, the interval distance between the welding wire end and the left side of the bevel and the interval distance between the welding wire end and the right side of the bevel are calculated respectively. Then, the difference between the two interval distances is calculated and the absolute value is taken to obtain the side wall spacing difference. Based on the coordinate data of the welding torch end center and the coordinate data of the welding trajectory centerline, the interval distance between the center of the welding torch end and the welding trajectory centerline is calculated to obtain the welding torch centering deviation. The arc spot image data is processed by grayscale threshold segmentation and connected region centroid extraction algorithm to extract the center position of the arc bright area. Then, the position difference is calculated with the welding trajectory centerline coordinate data to obtain the arc center offset. The molten pool contour image data is processed by edge detection and contour region centroid extraction algorithm to extract the molten pool contour center. The position is calculated by comparing the position difference with the coordinate data of the welding trajectory centerline to obtain the offset of the molten pool center. The fluctuation amplitude of the welding current data, welding voltage data, wire feed speed data, and welding speed data within the current welding window is calculated and then processed by the sliding window fluctuation statistics and time-series correlation to obtain the working condition disturbance. The square of the difference between the sidewall spacing and the value of one is calculated to obtain the sidewall constraint amplification. The centering deviation of the welding torch is calculated and the value of one is calculated to obtain the centering smoothing. The result of the sidewall constraint amplification divided by the centering smoothing is calculated, and the natural logarithm is taken to obtain the sidewall constraint logarithmic term. The absolute value of the difference between the arc center offset and the molten pool center offset is calculated to obtain the thermal input response deviation. The result of the thermal input response deviation divided by the centering smoothing is calculated and the exponential function value is taken to obtain the thermal input deviation amplification term. The working condition disturbance is calculated and the value of one is taken to obtain the working condition disturbance adjustment term. The product of the sidewall constraint logarithmic term, the thermal input deviation amplification term, and the working condition disturbance adjustment term is calculated to obtain the arc edge constraint value.

[0012] Furthermore, the specific process for determining and outputting the arc deviation state is as follows: Real-time comparison of the arc deviation constraint value and the arc deviation constraint threshold, which includes a primary constraint threshold and a secondary constraint threshold: When the arc deviation constraint value is less than the secondary constraint threshold, the current execution state corresponding to the welding torch end center coordinate data, welding torch forward tilt angle data, welding torch backward drag angle data, and welding torch yaw angle data is maintained, and the equivalent heat input center reconstruction analysis is initiated; When the arc deviation constraint value is greater than or equal to the secondary constraint threshold and less than the primary constraint threshold, an edge compensation flag is output, and the welding trajectory centerline coordinate data corresponding to the current welding section is used as a control reference and remains unchanged. Keep the current output state corresponding to the welding current data and welding voltage data unchanged, and send the side wall spacing difference, arc center offset, molten pool center offset, and welding torch centering deviation to the equivalent heat input center reconstruction analysis; when the arc edge constraint value is greater than or equal to the first-level constraint threshold, output the enhancement compensation flag, limit the swing amplitude of the welding torch yaw angle to the swing amplitude corresponding to the current enhancement compensation trigger time, freeze the current process state corresponding to the welding current data, welding voltage data, wire feed speed data, and welding speed data, and send the side wall spacing difference, arc center offset, molten pool center offset, welding torch centering deviation, and operating condition disturbance to the equivalent heat input center reconstruction analysis.

[0013] Furthermore, the specific process of performing equivalent heat input center reconstruction analysis on the original observation data and preceding control record data is as follows: The welding torch end center coordinate data, welding torch forward tilt angle data, welding torch backward drag angle data, welding torch yaw angle data, and welding wire end spatial coordinate data are subjected to attitude unfolding to obtain the incident position of the welding wire end relative to the welding trajectory centerline coordinate data; the surface contour point data of the previous layer of weld bead are subjected to surface unfolding corresponding to the current welding section to obtain the surface incident reference; the historical archive execution results that match the current welding torch circumferential position angle data in the preceding control record data are filtered to obtain the historical compensation reference results; the difference between the incident position and the surface incident reference is calculated to obtain the incident position deviation. The following steps are performed: 1. Calculate the difference between the arc center offset and the molten pool center offset to obtain the thermal input response deviation. 2. Calculate the difference between the sidewall spacing difference and the welding torch centering deviation to obtain the sidewall constraint mismatch. 3. Calculate the difference between the candidate adjustment results corresponding to the current welding torch end center coordinate data, welding torch forward tilt angle data, welding torch backward drag angle data, and welding torch yaw angle data, and the historical compensation reference results to obtain the historical compensation deviation. 4. Determine the consistency of the change direction and reduction magnitude of the incident position deviation, thermal input response deviation, sidewall constraint mismatch, and historical compensation deviation. 5. Map the number of consistent items obtained from the determination to the reduction magnitude to obtain the reliable value of the thermal input center reconstruction.

[0014] Furthermore, the specific process of correcting the welding torch posture and regressing the heat input region based on the equivalent heat input center reconstruction analysis results is as follows: Real-time comparison of the heat input center reconstruction confidence value and the heat input center reconstruction confidence threshold; when the heat input center reconstruction confidence value is less than the heat input center reconstruction confidence threshold, using the welding trajectory centerline coordinate data as a constraint benchmark, performing reverse deviation correction on the welding torch yaw angle data; using the welding torch end center coordinate data as the control object, performing lateral micro-displacement correction on the welding torch end center coordinate data; and using the welding wire end spatial coordinate data and the previous layer weld bead surface contour point data... Based on this, the forward tilt angle data and backward drag angle data of the welding torch are corrected in a linked manner; the archived execution results of the current circumferential position angle data of the welding torch in the preceding control record data are limited and called, and the corrected welding torch attitude correction results are sent to the convergence control analysis; when the heat input center reconstruction confidence value is greater than or equal to the heat input center reconstruction confidence threshold, a convergence reconstruction flag is output, the candidate adjustment results are kept from being added to the direction switch, the welding torch end center coordinate data are updated with a single position adjustment amount, and the candidate adjustment results are sent to the convergence control analysis as the current welding torch attitude correction results.

[0015] Furthermore, the specific process of executing control and performing convergence control analysis based on the welding torch posture correction results is as follows: Based on the welding torch posture correction results, control is executed on the welding torch end-center coordinate data, welding torch forward tilt angle data, welding torch backward drag angle data, and welding torch yaw angle data; after control execution, the arc center position is re-extracted from the arc spot image data, and the molten pool center position is re-extracted from the molten pool contour image data; the re-extracted arc center position and the re-extracted molten pool center position are aligned with the welding trajectory centerline coordinate data to obtain the compensated position offset state; then, the position offset is calculated for n consecutive welding windows. The compensated position offset state is continuously processed to obtain the regression continuity judgment result; the execution result matching the current layer number data and the welding torch circumferential position angle data is extracted from the previous control record data to obtain the historical backtracking reference result; the position difference between the compensated position offset state and the welding trajectory centerline coordinate data is calculated to obtain the offset change; then, the regression continuity judgment result and the historical backtracking reference result are combined to compare the consistency of the regression direction and the reduction magnitude of the offset change, and the convergence control value is obtained by using the number of consistent items and the reduction magnitude of the offset change as input.

[0016] Furthermore, the specific process of closed-loop feedback control and weld holding control based on the convergence control analysis results is as follows: Real-time comparison of convergence control value and convergence control threshold: When the convergence control value is less than the convergence control threshold, the current output state corresponding to the welding current data and welding voltage data remains unchanged, the angle amplitude corresponding to the welding torch sway angle data is restricted from continuing to change, and the current welding torch posture correction result is fed back to the equivalent heat input center for reconstruction analysis. The welding torch posture correction result is regenerated within the interval corresponding to the welding torch circumferential position angle data with the same index granularity as the previous control record database. When the convergence control value is greater than or equal to the convergence control threshold, a weld holding mark is output, the current welding torch posture correction result and the current process execution state are maintained, and the current layer number data, welding torch circumferential position angle data, current welding torch posture correction result, and corresponding welding current data, welding voltage data, wire feed speed data, and welding speed data are archived to the previous control record database.

[0017] The second aspect of this invention provides an adaptive control system for welding torch posture based on ultra-thick-walled pipe welding, comprising: an acquisition and preprocessing module for acquiring raw observation data during the operation of the welding equipment, obtaining preceding control record data, and preprocessing the raw observation data and preceding control record data; an asymmetric sidewall constraint discrimination module for jointly identifying the spatial constraint differences on both sides of the bevel, arc center offset characteristics, and molten pool response offset characteristics based on the raw observation data, and performing edge state discrimination and compensation level output; an equivalent heat input center reconstruction module for performing equivalent heat input center reconstruction analysis on the raw observation data and preceding control record data, and performing welding torch posture correction and heat input region regression control based on the equivalent heat input center reconstruction analysis results; and a closed-loop verification and stable control module for executing control based on the welding torch posture correction results and performing convergence control analysis, and performing closed-loop feedback control and weld holding control based on the convergence control analysis results. Attached Figure Description

[0018] Figure 1 This is a flowchart of the welding torch attitude adaptive control method based on ultra-thick wall pipe welding of the present invention;

[0019] Figure 2 This is a structural diagram of the welding torch attitude adaptive control system based on ultra-thick wall pipe welding of the present invention;

[0020] Figure 3 This is a bar chart showing the arc offset constraint values ​​of the present invention.

[0021] Figure 4 This is a three-dimensional relationship diagram of the sidewall spacing difference, thermal input response deviation, and arc deviation constraint value of the present invention.

[0022] Figure 5 The flowchart shows the closed-loop control process for reconstructing the reliable value of the heat input center in this invention. Detailed Implementation

[0023] To enable those skilled in the art to better understand the technical solution, the present invention will be described in detail below with reference to embodiments. The description in this part is only exemplary and explanatory, and should not be used to limit the scope of protection of the present invention in any way.

[0024] Please see Figures 1-5 This invention provides a technical solution: an adaptive control method for welding torch posture based on ultra-thick-walled pipe welding, comprising the following steps: S1, collecting raw observation data during the operation of the welding equipment, acquiring preceding control record data, and preprocessing the raw observation data and preceding control record data; S2, jointly identifying the spatial constraint differences on both sides of the bevel, arc center offset characteristics, and molten pool response offset characteristics based on the raw observation data, and performing edge state discrimination and compensation level output; S3, performing equivalent heat input center reconstruction analysis on the raw observation data and preceding control record data, and performing welding torch posture correction and heat input area regression control based on the equivalent heat input center reconstruction analysis results; S4, executing control based on the welding torch posture correction results and performing convergence control analysis, and performing closed-loop feedback control and weld holding control based on the convergence control analysis results.

[0025] Specifically, the process of collecting raw observation data during the operation of the welding equipment and obtaining preceding control record data is as follows: During the operation of the welding equipment, the raw observation data corresponding to the current welding section is received and organized. The current welding section is the instantaneous observation section when the welding torch moves along the center line of the welding trajectory to the corresponding sampling position. It is used to uniformly constrain the time correspondence of various spatial position data, attitude data, and image data. The raw observation data includes: left side wall contour point coordinate data of the bevel, right side wall contour point coordinate data of the bevel, center line coordinate data of the welding trajectory, center coordinate data of the welding torch end, forward tilt angle data of the welding torch, backward drag angle data of the welding torch, lateral sway angle data of the welding torch, spatial coordinate data of the welding wire end, and arc spot image data. The data includes weld pool contour image data, welding current data, welding voltage data, wire feed speed data, welding speed data, welding torch circumferential position angle data, current layer pass number data, and previous layer weld surface contour point data. Among these, the coordinate data of the left and right wall contour points of the bevel, the center line coordinates of the welding trajectory, the center coordinates of the welding torch tip, and the spatial coordinates of the welding wire tip all use a coordinate expression method under a unified workpiece coordinate system, with coordinate values ​​in millimeters. The angle units for the welding torch forward tilt angle, backward drag angle, lateral sway angle, and circumferential position angle are degrees. The coordinate data of the left and right wall contour points of the bevel and the center line coordinates of the welding trajectory are obtained from the weld seam... The tracking observation unit acquired data including the welding torch tip center coordinates, welding torch forward tilt angle, welding torch backward drag angle, welding torch yaw angle, and wire tip spatial coordinates, which were processed from the welding actuator's posture feedback data. Welding current, welding voltage, wire feed speed, and welding speed data were obtained from the welding controller's real-time output data. Arc spot image data and molten pool contour image data were acquired by the welding vision acquisition unit using a filtered imaging method. All types of raw observation data were processed at the same observation time at the same sampling location to ensure consistency in the data input used for subsequent spatial position determination, attitude determination, and process state determination. The left side wall contour point coordinates and the right side wall contour of the bevel were also collected. Point coordinate data, arc spot image data, and molten pool contour image data are used to characterize the spatial morphology and thermal state of the current welding section. Welding torch end center coordinate data, welding torch forward tilt angle data, welding torch backward drag angle data, welding torch lateral swing angle data, and wire end spatial coordinate data are used to characterize the current execution posture and incident state of the welding torch. Welding current data, welding voltage data, wire feed speed data, and welding speed data are used to characterize the current welding process state. The previous layer refers to the welding layer above the welding layer corresponding to the current layer number data. The surface contour point data of the previous layer weld bead is used to characterize the surface geometric boundary of the weld bead already formed below the current welding layer, providing a basis for judging the relationship between the actual incident position of the subsequent wire end and the surface incident reference.When the current layer number data corresponds to the first welding layer, the surface contour point data of the weld bead corresponding to the previous welding layer is empty, and the relevant processing based on the surface contour point data of the weld bead corresponding to the previous welding layer is skipped. This null value discrimination mechanism avoids the introduction of invalid surface reference information in the first layer welding scenario, and ensures that the data entry of the subsequent processing link remains consistent under different layer conditions.

[0026] Acquire preceding control record data and establish a preceding control record database. This database is used to associate and store data on the circumferential position angles of different welding torches, different pass / layer numbers, and corresponding execution results, supporting the extraction of subsequent historical compensation and backtracking reference results. The preceding control record data is obtained directly from the welding controller's historical execution log and is used to characterize the historical attitude and process execution results under the same welding torch circumferential position angle data and adjacent passes. The historical attitude execution results include historical control outputs corresponding to welding torch end-center coordinate data, welding torch forward tilt angle data, welding torch backward drag angle data, and welding torch yaw angle data. The process execution results include historical process states corresponding to welding current data, welding voltage data, wire feed speed data, and welding speed data.

[0027] In this implementation plan, this step achieves unified collection, organization, and entry management of spatial topography, welding torch posture, welding process status, and interlayer surface reference information under the current welding section. This enables the multi-source observation results of the current welding section to form a complete expression under the same data benchmark and maintain consistent processing boundaries under both first-layer and non-first-layer welding conditions. Simultaneously, by establishing a pre-control record database, the current welding status is correlated with historical posture execution results and process execution results under the same welding torch circumferential position angle data and adjacent layer number data. This provides a stable data support foundation for subsequent extraction of historical compensation reference results, retrieval of historical backtracking reference results, and continuous correction of welding torch posture adaptive control.

[0028] Specifically, the preprocessing of the raw observation data and preceding control record data involves: unifying the timestamp alignment of the raw observation data; mapping the coordinates of the left and right wall contour points of the bevel, the center line coordinates of the welding trajectory, the center coordinates of the welding torch tip, the spatial coordinates of the welding wire tip, the arc spot image data, the molten pool contour image data, as well as the welding current, welding voltage, wire feed speed, and welding speed data to the same sampling time to ensure a consistent time reference for subsequent spatial relationship calculations and process state determination. Invalid frame removal, missing frame completion, and continuous frame smoothing are performed on the arc spot image data and molten pool contour image data to reduce the impact of arc light obstruction, spatter adhesion, smoke disturbance, and instantaneous image distortion on the extraction results of the arc center position and molten pool center position, thereby improving the usability of the image sequence in continuous welding conditions. The surface contour point data of the previous layer of weld and the preceding control record data are processed by matching and organizing their circumferential positions. This establishes a positional association between the corresponding intervals of the previous layer of weld surface contour point data and the current welding torch circumferential position angle data. Simultaneously, the preceding control record data forms a retrievable historical reference under the same circumferential position and adjacent layer conditions. The original observation data and preceding control record data are standardized using a maximum absolute value standardization algorithm. By using the upper bound of the absolute value of each data sequence as a scaling benchmark, the dimensional differences between different physical quantities are eliminated, and the sign direction of each data is kept consistent before and after processing. The upper bound of the absolute value is statistically obtained from the corresponding data sequence after unified timestamp alignment and the corresponding historical reference sequence after circumferential position matching. This ensures that the standardization benchmark originates from the observation data corresponding to the current welding section and its retrievable historical reference data. Through a maximum-minimum normalization algorithm, the coordinate data of the left and right wall contour points of the bevel, the centerline coordinates of the welding trajectory, the center coordinates of the welding torch tip, the forward tilt angle, the backward drag angle, the lateral sway angle, the spatial coordinates of the welding wire tip, and the circumferential position angle of the welding torch are normalized. This ensures that various spatial position and attitude data are expressed within a unified numerical range, facilitating subsequent difference calculations, offset discrimination, and convergence comparisons. After standardization and normalization, all input quantities participating in subsequent natural logarithmic and exponential function calculations are dimensionless, thus providing a unified numerical basis for the sidewall spacing difference, welding torch centering deviation, arc center offset, molten pool center offset, operating condition disturbance, reliable value of heat input center reconstruction, and convergence control value.

[0029] In this implementation plan, this step achieves unified preprocessing and unified numerical expression of the original observation data and the preceding control record data, enabling the spatial position information, attitude information, image information, process information, and historical reference information under the current welding section to form a stable correspondence under the same time reference, the same position reference, and the same calculation scale. This improves the extractability and comparability of the arc center position, the molten pool center position, and related offset states, and provides a consistent, continuous, and directly callable data foundation for subsequent calculations of sidewall spacing difference, welding torch centering deviation, arc center offset, molten pool center offset, operating condition disturbance, heat input center reconstruction reliability value, and convergence control value.

[0030] Specifically, the process of jointly identifying the spatial constraint differences, arc center offset characteristics, and molten pool response offset characteristics on both sides of the bevel based on the original observation data is as follows: Based on the coordinate data of the contour points of the left and right sides of the bevel and the spatial coordinate data of the welding wire end, the interval distance between the welding wire end and the left side of the bevel and the interval distance between the welding wire end and the right side of the bevel are calculated respectively. Then, the difference between the two interval distances is calculated and the absolute value is taken to obtain the side wall spacing difference. The interval distance is calculated by the shortest point-to-line distance. Specifically, a side wall contour polyline is constructed based on the side wall contour point coordinate data. The point-to-line distance from the corresponding spatial point to each polyline segment is calculated, and the minimum distance is taken as the interval distance. By using the welding wire end as the spatial reference point of the current heat input position, the constraint relationship on both sides of the bevel can be transformed into a quantifiable cross-sectional imbalance characterization result. Based on the coordinate data of the welding torch tip center and the coordinate data of the welding trajectory centerline, the distance between the center of the welding torch tip and the centerline of the welding trajectory is calculated to obtain the welding torch centering deviation. This deviation is used to characterize the degree of deviation of the welding torch's geometric centering state from the nominal welding path, providing a basic reference for distinguishing between geometric offset and thermal input offset. The center position of the arc bright area is extracted from the arc spot image data using grayscale thresholding and connected region centroid extraction algorithms. The position difference is then calculated by subtracting the extracted center position coordinates from the corresponding reference point coordinates of the welding trajectory centerline in the horizontal coordinate component. The resulting difference is used as the corresponding offset, thus obtaining the arc center offset. Grayscale thresholding is used to separate the bright arc area from the arc spot image data, and connected region centroid extraction is used to determine the center coordinates of the bright arc area, thereby forming a stable representation of the arc thermal input center position. The center position of the molten pool contour is extracted from the molten pool contour image data using edge detection and contour region centroid extraction algorithms. The position difference is then calculated between this extracted position and the coordinate data of the welding trajectory centerline to obtain the molten pool center offset. Edge detection identifies the molten pool contour boundary, and contour region centroid extraction determines the center coordinates of the molten pool response area, thus reflecting the overall offset of the molten pool shape relative to the welding trajectory. The fluctuation amplitude of the welding current, welding voltage, wire feed speed, and welding speed data is calculated separately within the current welding window. This fluctuation amplitude is then fused using sliding window fluctuation statistics and time-series correlation analysis to obtain the operating condition disturbance. The sliding window fluctuation statistics and time-series correlation analysis involves statistically analyzing the local fluctuation amplitude of each data point within the current welding window and combining this with the consistency of the data's change direction and the correlation of the fluctuation process to obtain the operating condition disturbance.The current welding window is formed by extracting consecutive welding sampling points before and after the current sampling time corresponding to the circumferential position angle data of the welding torch. The current welding window is defined by the real-time sampling sequence of the welding controller, and the window duration is 0.2 seconds to 1.5 seconds. Then, through the fusion processing of sliding window fluctuation statistics and time-series correlation, the synchronization is first judged according to the consistency of the fluctuation direction of each data within the same current welding window, and then the coupling is judged according to the overlap of the fluctuation amplitude of each data on the time axis. Finally, the synchronization judgment result and the coupling judgment result are mapped into a single disturbance characterization result to characterize the joint disturbance effect of the current welding condition on the arc center position and the molten pool center position.

[0031] The square of the sidewall spacing difference plus one is calculated to obtain the sidewall constraint amplification. This square mapping enhances the amplitude resolution of the sidewall spacing difference under asymmetric bevel conditions, providing a clearer amplification representation of the sidewall spatial constraint difference in subsequent discrimination. The centering deviation of the welding torch is calculated plus one to obtain the centering smoothing amount. This addition suppresses the sensitive abrupt change in the denominator when the welding torch centering deviation approaches zero, ensuring continuous and bounded ratio calculations in subsequent calculations. The result of dividing the sidewall constraint amplification by the centering smoothing amount is calculated, then the natural logarithm is taken after adding one to obtain the sidewall constraint logarithmic term. Natural logarithmic compression controls the growth slope of the sidewall constraint amplification within a large fluctuation range, maintaining comparability of constraint differences between different welding sections. The absolute value of the difference between the arc center offset and the molten pool center offset is calculated to obtain the thermal input response deviation. This absolute value form uniformly represents the degree of response mismatch between the arc center position and the molten pool center position, avoiding the cancellation of deviation intensity judgments due to differences in offset direction. The thermal input response deviation is calculated by dividing the result by the centering smoothing factor, and then the exponential function value is taken to obtain the thermal input deviation amplification term. This exponential mapping highlights the abnormal sensitivity of the thermal input response deviation under conditions of small welding torch centering deviation, thus strengthening the expression of the inconsistency between the thermal input center and the molten pool response center. The independent variable of the exponential function is limited to zero to α, where α is the upper bound control parameter of the exponential mapping, determined based on the historical statistical distribution of the thermal input response deviation and welding torch centering deviation under real-time sampling conditions of the welding equipment. This parameter limits the exponential amplification intensity and has a value range of 1 to 3. The operating condition disturbance is calculated and then incremented by one to obtain the operating condition disturbance adjustment term. Through translation, the operating condition disturbance participates in the overall discrimination process as a multiplicative adjustment factor, thereby reflecting the additional influence of the combined fluctuations of welding current data, welding voltage data, wire feed speed data, and welding speed data on the edge deviation state. The arc deviation constraint value is obtained by multiplying the logarithmic term of the sidewall constraint, the thermal input deviation amplification term, and the operating condition disturbance adjustment term. A joint discrimination result is constructed from three dimensions: spatial constraint, thermal input response, and operating condition disturbance. This allows the arc deviation constraint value to comprehensively reflect the unilateral traction risk of the current welding section. All the above quantities are dimensionless quantities normalized to zero to one. The arc deviation constraint value is a dimensionless discrimination value. The specific calculation formula is as follows:

[0032] ;

[0033] In the formula: It represents the arc offset constraint value; it is used to characterize the offset constraint strength under the combined effects of asymmetric spatial constraints on both sides of the bevel, arc center position offset, molten pool center position offset, and working condition disturbances. It represents the difference in sidewall spacing; it is used to characterize the degree of unevenness in the spatial spacing between the welding wire tip and the left and right walls of the bevel. It indicates the centering deviation of the welding torch; it is used to characterize the degree of deviation of the center of the welding torch tip from the center line of the welding trajectory. It represents the offset of the arc center; it is used to characterize the lateral offset of the arc heat input center relative to the welding trajectory centerline. It represents the offset of the molten pool center; it is used to characterize the offset state of the molten pool response center relative to the welding trajectory centerline. It represents the amount of disturbance in the working condition; it is used to characterize the degree of combined disturbance of the welding process state fluctuations within the current welding window to the arc center position and the molten pool center position.

[0034] In this embodiment, Table 1 is an example data table of arc deviation constraint values, which records in detail the difference in sidewall spacing, welding torch centering deviation, arc center offset, molten pool center offset, working condition disturbance, and the final calculated arc deviation constraint values ​​for different welding sections during the analysis process. This is used to quantify the influence of asymmetric constraints on both sides of the bevel on the arc heat input center offset under different welding sections. Wherein: the sidewall spacing difference for welding section T1 is 0.08, the welding torch centering deviation is 0.05, the arc center offset is 0.06, the molten pool center offset is 0.05, the working condition disturbance is 0.04, and the arc edge constraint value is 0.163; the sidewall spacing difference for welding section T2 is 0.15, the welding torch centering deviation is 0.07, the arc center offset is 0.12, the molten pool center offset is 0.09, the working condition disturbance is 0.06, and the arc edge constraint value is 0.274; the sidewall spacing difference for welding section T3 is 0.26, the welding torch centering deviation is 0.10, and the arc center offset is... The value is 0.21, the molten pool center offset is 0.15, the working condition disturbance is 0.09, and the arc edge constraint value is 0.460; the side wall spacing difference corresponding to welding section T4 is 0.39, the welding torch centering deviation is 0.12, the arc center offset is 0.31, the molten pool center offset is 0.20, the working condition disturbance is 0.13, and the arc edge constraint value is 0.713; the side wall spacing difference corresponding to welding section T5 is 0.52, the welding torch centering deviation is 0.16, the arc center offset is 0.43, the molten pool center offset is 0.26, the working condition disturbance is 0.18, and the arc edge constraint value is 1.064.

[0035] Table 1 Example Data Table of Arc Offset Constraint Values

[0036]

[0037] like Figure 3The bar chart shown is a histogram of arc deviation constraint values. Combined with Table 1, it can be seen that the arc deviation constraint values ​​for different welding sections show a gradual upward trend. Among them, welding section T5 has the highest arc deviation constraint value at 1.064, indicating that the combined effects of asymmetric constraints on both sides of the bevel, arc center offset, and operating condition disturbances are most significant at this welding section. Welding section T1 has the lowest arc deviation constraint value at only 0.163, indicating that the arc heat input center offset is relatively weak at this welding section. The arc deviation constraint values ​​for welding sections T2, T3, and T4 are 0.274, 0.460, and 0.713 respectively, showing that the arc deviation constraint value continuously increases with the increase of the sidewall spacing difference, arc center offset, and operating condition disturbance. Overall, the histogram of arc deviation constraint values ​​intuitively reflects the differences in the deviation constraint strength of different welding sections and can serve as a basis for deviation state judgment and compensation level output.

[0038] like Figure 4 The figure shows a three-dimensional relationship between the sidewall spacing difference, thermal input response deviation, and arc edge constraint value. Combined with Table 1, it can be seen that the arc edge constraint value increases overall with the increase of both the sidewall spacing difference and the thermal input response deviation. The three-dimensional surface reflects the combined influence of these two factors on the arc edge constraint value. The red measured points correspond to the actual data distribution of welding sections T1 to T5. Welding sections T1 and T2 are located in the lower region of the surface, indicating that when both the sidewall spacing difference and the thermal input response deviation are small, the arc edge constraint value remains at a low level. Welding section T3 is in the intermediate transition region, indicating that the edge constraint has begun to strengthen. Welding sections T4 and T5 are located in the higher region of the surface, showing that under the conditions of increased sidewall spacing difference and a widening difference between the arc center offset and the molten pool center offset, the arc edge constraint value is significantly amplified. Overall, the three-dimensional relationship diagram can intuitively reveal the coupling relationship between the sidewall asymmetric constraint and the thermal input response deviation on the arc edge constraint value, and can serve as a visual basis for verifying the formula change trend and illustrating the embodiments.

[0039] In this implementation plan, this step achieves a joint quantitative characterization of the constraint states on both sides of the current welding section's bevel, the welding torch's geometric alignment state, the arc heat input center offset state, the molten pool response center offset state, and the welding process fluctuation state. Based on this, an arc deviation constraint value is constructed, further expanding the discrimination method that relies solely on the welding torch's geometric alignment into a comprehensive discrimination method that addresses the superimposed effects of asymmetric sidewall traction, heat input response mismatch, and operating condition disturbances. This enables a more accurate identification of whether there is a risk of unilateral traction in the current welding section and its intensity, providing a clear, stable, and graded discrimination basis for subsequent equivalent heat input center reconstruction analysis.

[0040] Specifically, the process of determining the edge state and outputting the compensation level is as follows: Real-time comparison of the arc edge constraint value and the arc edge constraint threshold. The arc edge constraint threshold includes a first-level constraint threshold and a second-level constraint threshold. A hierarchical threshold discrimination method is used to distinguish the degree of unilateral constraint traction, so that different edge states correspond to different levels of compensation entry.

[0041] When the arc deviation constraint value is less than the secondary constraint threshold, it is determined that the current welding section does not form a significant unilateral constraint pull. The current execution state corresponding to the welding torch end center coordinate data, welding torch forward tilt angle data, welding torch backward drag angle data, and welding torch yaw angle data is maintained, and the equivalent heat input center reconstruction analysis is entered. At this time, the current execution state is retained to avoid introducing additional attitude fluctuations under slight disturbance conditions, so that the subsequent equivalent heat input center reconstruction analysis can continue on the basis of the existing working conditions.

[0042] When the arc deviation constraint value is greater than or equal to the secondary constraint threshold and less than the primary constraint threshold, it is determined that an arc deviation trend has been formed but has not yet entered a continuous traction state. An deviation compensation mark is output, and the coordinate data of the welding trajectory centerline corresponding to the current welding section is used as a control reference and kept unchanged. The current output state corresponding to the welding current data and welding voltage data remains unchanged. The difference in sidewall spacing, arc center offset, molten pool center offset, and welding torch alignment deviation are sent to the equivalent heat input center reconstruction analysis. By locking the current welding section, subsequent attitude corrections are limited to the same spatial reference to avoid drift of the discrimination benchmark due to trajectory switching during the compensation process.

[0043] When the arc offset constraint value is greater than or equal to the first-level constraint threshold, it is determined that the current asymmetric constraint of the sidewall has caused a continuous offset effect on the arc center position and the molten pool center position. An enhanced compensation flag is output, limiting the swing amplitude of the welding torch yaw angle to the amplitude corresponding to the current enhanced compensation trigger moment, preventing further incremental changes along the outward expansion direction. The current process state corresponding to the welding current data, welding voltage data, wire feed speed data, and welding speed data is frozen. The sidewall spacing difference, arc center offset, molten pool center offset, welding torch alignment deviation, and operating condition disturbance are sent to the equivalent heat input center reconstruction analysis. By freezing the current process state, subsequent compensation is prioritized for the welding torch pose correction process, avoiding superimposed disturbances caused by synchronous changes in process parameters in the determination of the arc center position and the molten pool center position.

[0044] In this implementation scheme, this step achieves graded discrimination and layered control of the arc deviation constraint value, enabling the single-sided constraint traction state of the current welding section to be accurately distinguished according to the degree of deviation formation, and corresponding to different levels of control entry points such as welding torch posture maintenance, deviation compensation intervention, and enhanced compensation triggering. On this basis, the adjustment boundaries of the welding torch end center coordinate data, welding torch forward tilt angle data, welding torch backward drag angle data, welding torch yaw angle data, as well as welding current data, welding voltage data, wire feed speed data, and welding speed data are subject to targeted constraints. This ensures that the subsequent equivalent heat input center reconstruction analysis is always conducted under control conditions that match the current risk intensity, reducing unnecessary posture fluctuations under slight disturbance conditions, suppressing additional interference caused by synchronous changes in process parameters under continuous deviation conditions, and improving the stability, pertinence, and continuity of the welding torch posture correction process.

[0045] Specifically, the process of reconstructing the equivalent heat input center from the original observation data and the preceding control record data is as follows: Attitude unfolding is performed on the welding torch tip center coordinate data, welding torch forward tilt angle data, welding torch backward drag angle data, welding torch yaw angle data, and welding wire tip spatial coordinate data. Attitude unfolding involves constructing the spatial orientation relationship of the welding wire tip based on the welding torch tip center coordinate data, welding torch forward tilt angle data, welding torch backward drag angle data, and welding torch yaw angle data. The incident position of the welding wire tip relative to the welding trajectory centerline coordinate data is then calculated by combining the welding wire tip spatial coordinate data. By uniformly unfolding the welding torch pose information and the welding wire tip spatial position, the changes in the welding torch spatial attitude can be transformed into a representation of the actual position of the welding wire tip relative to the welding trajectory centerline coordinate data. The surface contour point data of the previous layer of weld bead is used to unfold the surface corresponding to the current welding section. Surface unfolding involves contour interpolation based on the previous layer's weld bead surface contour point data, and then projecting the interpolated surface contour onto the current welding section to obtain the surface incidence reference. This reference is used to establish a reference relationship between the incident direction of the welding wire tip under the current layer and the surface morphology of the already formed weld bead below. Historical archived execution results matching the current welding torch circumferential position angle data in the preceding control record data are filtered to obtain historical compensation reference results. By extracting archived execution results with the same welding torch circumferential position angle data, an inheritable historical reference is provided for the current welding torch attitude correction. The difference between the incident position and the surface incidence reference is calculated to obtain the incident position deviation, which characterizes the degree of deviation of the current incident position of the welding wire tip relative to the target surface incident position. The difference between the arc center offset and the molten pool center offset is calculated to obtain the heat input response deviation, which characterizes the consistency between the arc heat input center and the molten pool response center. The difference between the sidewall spacing difference and the welding torch centering deviation is calculated to obtain the sidewall constraint mismatch; this is used to characterize the degree of adaptation between the spatial constraint changes on both sides of the bevel and the geometric centering state of the welding torch. The difference between the candidate adjustment results corresponding to the current welding torch end center coordinate data, welding torch forward tilt angle data, welding torch backward drag angle data, and welding torch yaw angle data and the historical compensation reference results is calculated to obtain the historical compensation deviation. The candidate adjustment results are the adjustment results to be executed corresponding to the current welding torch end center coordinate data, welding torch forward tilt angle data, welding torch backward drag angle data, and welding torch yaw angle data. By comparing the difference between the current adjustment results to be executed and the historical compensation reference results, the current welding torch attitude correction is constrained not to deviate from the existing archived execution range.

[0046] The consistency of the changing direction and the consistency of the reduction magnitude of the incident position deviation, thermal input response deviation, sidewall constraint mismatch, and historical compensation deviation are judged. The number of consistent items and the reduction magnitude are comprehensively mapped to obtain the thermal input center reconstruction confidence value, which is used to characterize the common regression ability of the welding torch posture correction result generated based on the current edge state to the actual incident position of the welding wire end, the arc center position, and the molten pool center position. Among them, the consistency of the changing direction is used to judge whether the deviations evolve together in the direction of reduction, and the consistency of the reduction magnitude is used to judge whether the convergence rhythm of the deviations is coordinated, so that the thermal input center reconstruction confidence value can comprehensively reflect the regression ability of the current candidate adjustment result to the target thermal input region. The target thermal input region is the thermal input action range determined based on the welding trajectory centerline coordinate data, the surface contour point data of the previous layer of weld, and the sidewall spatial relationship of the current welding section.

[0047] In this implementation plan, this step achieves joint expansion and collaborative verification of the welding torch spatial attitude, the actual incident position of the welding wire tip, the surface morphology of the previous weld bead, and the historical archived execution results. This transforms the welding torch attitude correction under the current welding section from a single position offset compensation to a comprehensive regression and discrimination process around the target heat input region. By simultaneously introducing the incident position deviation, heat input response deviation, sidewall constraint mismatch, and historical compensation deviation, the welding torch attitude correction results can be uniformly evaluated from multiple levels, including the actual action position of the welding wire tip, the matching relationship between the arc heat input center and the molten pool response center, the adaptation relationship between the bevel sidewall constraint and the welding torch geometry alignment, and the inheritance relationship of the current candidate adjustment results to the historical effective execution range. This results in a reliable value for the reconstruction of the heat input center, providing a clear basis for subsequent judgment on whether the current candidate adjustment results meet the conditions for entering the convergence control stage. At the same time, it improves the welding torch attitude correction results' fit to the target heat input region, historical inheritance ability, and continuous regression ability.

[0048] Specifically, the process of correcting the welding torch posture and regressing the heat input region based on the equivalent heat input center reconstruction analysis results is as follows: Figure 5 The diagram shows the closed-loop control flowchart for the reconstructed reliability value of the heat input center. It compares the reconstructed reliability value and the reconstructed reliability threshold of the heat input center in real time, and determines whether the candidate adjustment results meet the conditions for entering the closed-loop verification stage, thus ensuring that the regression control of the heat input region maintains a clear convergence entry point.

[0049] When the confidence value of the heat input center reconstruction is less than the confidence threshold of the heat input center reconstruction, it is determined that the current welding torch posture correction result cannot make the actual incident position of the welding wire end, the arc center position, and the molten pool center position form a common regression relationship. Using the welding trajectory centerline coordinate data as a constraint benchmark, the welding torch yaw angle data is corrected in the opposite direction; the opposite direction correction is performed along the direction that reduces the arc center offset. Using the welding torch end center coordinate data as the control object, the welding torch end center coordinate data is corrected in the lateral micro-displacement direction; the correction magnitude of the lateral micro-displacement correction is determined based on the current magnitude of the arc center offset and the welding torch alignment deviation, and is limited to the historical archive execution range corresponding to the previous control record data. Based on the welding wire end spatial coordinate data and the previous layer of weld bead surface contour point data, the welding torch forward tilt angle data and welding torch backward drag angle data are corrected in a linked manner; and the corresponding data in the previous control record data are also corrected. The archived execution results of the current welding torch circumferential position angle data are subjected to amplitude-limited calls. Amplitude-limited calls involve extracting the historical control outputs corresponding to the current welding torch end-center coordinates, forward tilt angle, backward drag angle, and yaw angle from the preceding control record data. The range of variation in these historical control outputs is used as the amplitude constraint range for the current welding torch attitude correction result, ensuring that the welding torch attitude correction result remains within the amplitude range corresponding to the historical archived execution results. The corrected welding torch attitude correction result is then sent to the convergence control analysis. Specifically, the reverse bias correction is used to preferentially offset the unilateral offset of the current arc center position relative to the welding trajectory centerline; the lateral micro-displacement correction is used to compensate for the residual deviation of the welding torch end-center coordinates in the geometric alignment direction; and the linkage correction is used to synchronously adjust the relative relationship between the actual incident position of the welding wire end and the surface corresponding to the previous layer of weld bead surface contour points. By using the reconstructed reliability value of the heat input center as the basis for switching the current welding torch attitude correction path, and combining this with the amplitude-limited calls to the archived execution results corresponding to the preceding control record data, the continuity and stability of the welding torch attitude correction process are improved.

[0050] When the confidence value of the heat input center reconstruction is greater than or equal to the confidence threshold of the heat input center reconstruction, it is determined that the current welding torch posture correction result can establish a regression relationship of the target heat input area. A convergence reconstruction flag is output, and the candidate adjustment results are kept from further direction switching. A single position adjustment is performed on the welding torch end center coordinate data. The single position adjustment is the position increment within the current welding control cycle, and the candidate adjustment results are sent to the convergence control analysis as the current welding torch posture correction result. By stopping the addition of direction switching, the reverse control action is avoided after a regression trend has been formed. The single position adjustment update allows the welding torch end center coordinate data to gradually conform along the current regression direction.

[0051] In this implementation scheme, this step achieves threshold discrimination and directional adjustment of the reliability value of the heat input center reconstruction, enabling the current candidate adjustment results to be further processed according to whether they have the regression capability of the target heat input area. When the reliability value of the heat input center reconstruction is insufficient, targeted corrections are made to the welding torch yaw angle data, welding torch end center coordinate data, welding torch forward tilt angle data, and welding torch backward drag angle data. Combined with the archived execution results corresponding to the previous control record data, the amplitude is limited and called, so that the welding torch attitude correction results continue to converge towards the common regression direction of the actual incident position of the welding wire end, the arc center position, and the molten pool center position while maintaining the historical executable boundary. When the reliability value of the heat input center reconstruction reaches the reliability threshold of the heat input center reconstruction, the current candidate adjustment results are stably retained and progressively fitted along the current regression direction, so that the welding torch attitude correction process forms a clear switch between correction enhancement and direction maintenance, improving the stability, continuity, and convergence of the target heat input area regression control.

[0052] Specifically, the process of executing control and performing convergent control analysis based on the welding torch attitude correction results is as follows: Control is executed on the welding torch end-center coordinate data, welding torch forward tilt angle data, welding torch backward drag angle data, and welding torch yaw angle data according to the welding torch attitude correction results. By converting the welding torch attitude correction results into synchronous control quantities for the welding torch end-position and attitude angle, the spatial position adjustment and attitude direction adjustment of the welding torch are completed collaboratively within the same control cycle. After control execution, the arc center position is re-extracted from the arc spot image data, and the molten pool center position is re-extracted from the molten pool contour image data. This is used to re-acquire the actual distribution state of the arc heat input center and the molten pool response center after attitude correction, forming feedback observations oriented towards the execution results. The re-extracted arc center position and re-extracted molten pool center position are aligned with the welding trajectory centerline coordinate data to obtain the compensated position offset state. This ensures that the degree of offset between the arc center position and the molten pool center position relative to the welding trajectory centerline is uniformly characterized after compensation execution. The compensated position offset states within n consecutive welding windows are then processed for continuity, yielding a regression continuity judgment result. This result indicates that within n consecutive welding windows, the arc center position and the molten pool center position consistently revert to the welding trajectory centerline without any reverse jumps. n is determined by the number of sampling windows in the current welding control cycle and the number of consecutive regression windows corresponding to the same welding torch circumferential position angle data in the preceding control record data, with n ranging from 3 to 8. Introducing time-series discrimination across n consecutive welding windows avoids misjudging instantaneous regression phenomena within a single welding window as effective convergence. Execution results matching the current layer number data and welding torch circumferential position angle data are extracted from the preceding control record data to obtain historical backtracking reference results. These results provide a historical execution reference for the current welding section, constraining the accumulation of abnormal offsets during the closed-loop control process.

[0053] The positional difference between the compensated positional offset state and the coordinate data of the welding trajectory centerline is calculated to obtain the offset change; this is used to characterize the change in the arc center position and the molten pool center position relative to the welding trajectory centerline before and after compensation. Combining the continuous regression results and historical backtracking reference results, a consistency comparison is made between the regression direction and the magnitude of the reduction in offset change. Using the number of consistent terms and the magnitude of the reduction in offset change as inputs, a convergence control value is obtained. The convergence control value is a dimensionless quantity between zero and one, obtained by weighted mapping of the number of consistent terms and the magnitude of the reduction in offset change. The closer the convergence control value is to one, the stronger the convergence of the current welding torch posture correction result. The number of consistent terms is used to characterize the degree to which the arc center position and the molten pool center position maintain common regression within the continuous welding window. The magnitude of the reduction in offset change is used to characterize the effectiveness of the compensation execution in suppressing offset and to characterize the constraint effect of the welding torch posture correction result on the arc center position and the molten pool center position. It is determined whether the arc center position and the molten pool center position return to the allowable range and maintain the same regression direction within the continuous sampling window without reverse jump. Thus, the convergence control value can simultaneously reflect the instantaneous offset improvement and the continuous regression maintenance. The allowable range refers to the range of permissible offset between the arc center position and the molten pool center position, established based on the coordinate data of the welding trajectory centerline. The allowable range is used to limit the acceptable offset boundaries of the arc center position and the molten pool center position relative to the welding trajectory centerline, providing a unified judgment basis for subsequent closed-loop feedback control and welding hold control.

[0054] In this implementation plan, this step realizes the actual execution, feedback observation, and closed-loop convergence discrimination of the welding torch attitude correction results. This allows the adjustment results of the welding torch end center coordinate data, welding torch forward tilt angle data, welding torch backward drag angle data, and welding torch yaw angle data to be promptly converted into feedback verification of the changes in the arc center position and molten pool center position after control execution. It also allows for continuous tracking and unified evaluation of the compensated position offset state within the continuous welding window. Simultaneously, by introducing regression continuous discrimination results, historical backtracking reference results, and offset change amounts, the instantaneous offset improvement capability, continuous regression maintenance capability, and abnormal offset suppression capability of the current welding torch attitude correction results are comprehensively constrained. This enables the convergence control value to serve as a unified criterion for whether the welding torch attitude correction results have met the stable convergence requirements, thereby providing a stable, continuous, and historically referenced control foundation for subsequent closed-loop feedback control or welding hold control.

[0055] Specifically, the process of performing closed-loop feedback control and weld holding control based on the convergence control analysis results is as follows: Real-time comparison of the convergence control value and the convergence control threshold; unified judgment of the compensated position offset state and continuous regression state using the convergence control value; enabling the closed-loop control process to switch between feedback correction and holding output.

[0056] When the convergence control value is less than the convergence control threshold, it is determined that the current welding torch posture correction result has not reached the stable convergence requirement. The current output state corresponding to the welding current data and welding voltage data remains unchanged, the angle amplitude corresponding to the welding torch yaw angle data is restricted from continuing to change, and the current welding torch posture correction result is sent back to the equivalent heat input center for reconstruction analysis. The welding torch posture correction result is regenerated within the interval corresponding to the welding torch circumferential position angle data with the same index granularity as the previous control record database. The interval corresponding to the welding torch circumferential position angle data is the circumferential position angle interval formed by dividing by step size a. Step size a is determined by the circumferential travel coding resolution of the welding equipment and the position angle index interval of the previous control record database, and the value range is 1° to 10°. By sending it back to the equivalent heat input center for reconstruction analysis, the welding torch posture correction result is recalculated within the same interval corresponding to the welding torch circumferential position angle data, avoiding cross-interval correction that leads to compensation direction mismatch.

[0057] When the convergence control value is greater than or equal to the convergence control threshold, it is determined that the current welding torch posture correction result has met the convergence requirements. A welding hold flag is output, maintaining the current welding torch posture correction result and the current process execution state. The current pass number data, welding torch circumferential position angle data, current welding torch posture correction result, and corresponding welding current data, welding voltage data, wire feed speed data, and welding speed data are archived to the preceding control record database. This serves as the priority call result for subsequent calls to the same welding torch circumferential position angle data and adjacent passes. By archiving the execution results corresponding to the current pass number data and welding torch circumferential position angle data, historical data is provided for rapid recall and rollback control under the same operating conditions.

[0058] In this implementation scheme, this step achieves final threshold adjudication and closed-loop control diversion for the convergence control value. This allows the welding torch attitude correction result to be clearly divided into two control paths based on the current compensated position offset state and continuous regression state: either continue to send back correction or maintain stable output. When the convergence requirement is not met, by maintaining the current output state corresponding to the welding current data and welding voltage data and limiting the angle amplitude corresponding to the welding torch yaw angle data, the compensation process returns to the equivalent heat input center for reconstruction analysis within the controlled boundary. This ensures that the welding torch attitude correction result is always recalculated specifically around the interval corresponding to the current welding torch circumferential position angle data, reducing the risk of compensation direction drift and cross-interval mismatch. When the convergence requirement is met, by maintaining the current welding torch attitude correction result and the current process execution state, and archiving the corresponding execution result to the preceding control record database, the effective control result of the current welding section can be accumulated as the priority reference for subsequent use of the same welding torch circumferential position angle data and adjacent layer conditions. This improves the stability maintenance capability, historical reuse capability, and continuous control capability of the entire closed-loop control process.

[0059] like Figure 2 As shown, the second aspect of this invention provides an adaptive control system for welding torch posture based on ultra-thick-walled pipe welding, comprising: an acquisition and preprocessing module, used to acquire raw observation data during the operation of the welding equipment, obtain preceding control record data, and preprocess the raw observation data and preceding control record data; by forming a unified time reference, a unified position reference, and a unified numerical expression, it provides a continuous data foundation for subsequent edge discrimination, heat input center reconstruction, and closed-loop control. An asymmetric sidewall constraint discrimination module, used to jointly identify the spatial constraint differences on both sides of the bevel, arc center offset characteristics, and molten pool response offset characteristics based on the raw observation data, and to perform edge state discrimination and compensation level output; by classifying and identifying the risk of unilateral constraint traction and its degree of formation, it provides a clear compensation entry point for subsequent welding torch posture correction. An equivalent heat input center reconstruction module, used to perform equivalent heat input center reconstruction analysis on the raw observation data and preceding control record data, and to perform welding torch posture correction and heat input area regression control based on the equivalent heat input center reconstruction analysis results; by establishing a regression relationship around the target heat input area, it improves the fitting ability and historical inheritance ability of the welding torch posture correction results. The closed-loop verification and stabilization control module is used to execute control based on the welding torch attitude correction results and perform convergence control analysis. Based on the convergence control analysis results, it performs closed-loop feedback control and welding hold control. By performing feedback verification and convergence discrimination on the execution results, it achieves adaptive switching between feedback correction and stable holding of the welding torch attitude correction results.

[0060] In this implementation scheme, through the sequential collaboration of the acquisition and preprocessing module, the asymmetric sidewall constraint discrimination module, the equivalent heat input center reconstruction module, and the closed-loop verification and stability control module, continuous sensing, hierarchical discrimination, directional correction, and closed-loop convergence control of the asymmetric constraints on both sides of the bevel, the offset of the arc center position, the offset of the molten pool center position, and the results of the welding torch posture correction during the welding of ultra-thick-walled pipes are achieved. This enables the welding torch posture control to further shift from traditional geometric alignment adjustment to regression control oriented towards the target heat input area. Under the joint constraints of historical execution reference and real-time feedback verification, the pertinence, continuity, and stability of the welding torch posture correction are improved, thereby enhancing the heating consistency, fusion balance, and weld bead formation stability on both sides of the bevel.

[0061] It should be noted that, in this document, the terms "comprising," "including," and any other variations are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Specific examples have been used in this document to illustrate the principles and implementation methods of the present invention. These examples are merely for the purpose of helping to understand the method and core ideas of the present invention. The above descriptions are only preferred embodiments of the present invention. It should be pointed out that, due to the limitations of written expression and the objective existence of infinite specific structures, those skilled in the art can make several improvements, modifications, or variations without departing from the principles of the present invention, and can also combine the above technical features in an appropriate manner. These improvements, modifications, variations, or combinations, or the direct application of the concept and technical solution of the present invention to other situations without modification, should all be considered within the scope of protection of the present invention.

Claims

1. An adaptive control method for welding torch posture based on ultra-thick-walled pipe welding, characterized in that, Includes the following steps: S1, collect raw observation data during the operation of the welding equipment, obtain the preceding control record data, and preprocess the raw observation data and the preceding control record data; S2, based on the original observation data, jointly identifies the spatial constraint differences on both sides of the bevel, the characteristics of the arc center offset, and the characteristics of the molten pool response offset, and performs edge state discrimination and compensation level output. S3, perform equivalent heat input center reconstruction analysis on the original observation data and the preceding control record data, and perform welding torch posture correction and heat input area regression control based on the equivalent heat input center reconstruction analysis results; S4, based on the welding torch posture correction results, executes control and performs convergence control analysis, and performs closed-loop feedback control and welding hold control based on the convergence control analysis results.

2. The adaptive control method for welding torch posture based on ultra-thick-walled pipe welding according to claim 1, characterized in that: The specific process of acquiring the raw observation data during the operation of the welding equipment and obtaining the preceding control record data is as follows: During the operation of the welding equipment, the original observation data corresponding to the current welding section is received and processed. The original observation data includes: the coordinate data of the contour points of the left side wall of the bevel, the coordinate data of the contour points of the right side wall of the bevel, the coordinate data of the center line of the welding trajectory, the coordinate data of the center of the welding torch end, the forward tilt angle of the welding torch, the backward drag angle of the welding torch, the lateral swing angle of the welding torch, the spatial coordinate data of the welding wire end, the arc spot image data, the molten pool contour image data, the welding current data, the welding voltage data, the wire feed speed data, the welding speed data, the circumferential position angle data of the welding torch, the current layer pass number data, and the surface contour point data of the previous layer pass. The previous layer refers to the welding layer above the welding layer corresponding to the current layer pass number data. Acquire preceding control record data and establish a preceding control record database.

3. The adaptive control method for welding torch posture based on ultra-thick-walled pipe welding according to claim 1, characterized in that: The specific process for preprocessing the raw observation data and preceding control record data is as follows: The original observation data were aligned with a unified timestamp; invalid frames were removed, missing frames were filled, and continuous frames were smoothed for the arc spot image data and molten pool contour image data; the surface contour point data of the previous layer of weld and the preceding control record data were matched and processed in the same circumferential direction; the original observation data and the preceding control record data were standardized using the maximum absolute value standardization algorithm; and the coordinate data of the left side wall contour point of the bevel, the right side wall contour point of the bevel, the center line coordinate data of the welding trajectory, the center coordinate data of the welding torch end, the forward tilt angle of the welding torch, the backward drag angle of the welding torch, the lateral sway angle of the welding torch, the spatial coordinate data of the welding wire end, and the circumferential position angle data of the welding torch were normalized using the maximum and minimum value normalization algorithm.

4. The adaptive control method for welding torch posture based on ultra-thick-walled pipe welding according to claim 1, characterized in that: The specific process for jointly identifying the spatial constraint differences on both sides of the bevel, the arc center offset characteristics, and the molten pool response offset characteristics based on the original observation data is as follows: Based on the coordinate data of the contour points of the left and right sides of the bevel and the spatial coordinate data of the welding wire tip, the interval distance between the welding wire tip and the left side of the bevel and the interval distance between the welding wire tip and the right side of the bevel are calculated respectively. The difference between the two interval distances is then taken as the absolute value to obtain the side wall spacing difference. Based on the coordinate data of the welding torch tip center and the coordinate data of the welding trajectory centerline, the interval distance between the center of the welding torch tip and the welding trajectory centerline is calculated to obtain the welding torch centering deviation. The arc spot image data is processed by grayscale threshold segmentation and connected region centroid extraction algorithm to extract the center position of the arc highlight area. The position difference is then calculated with the welding trajectory centerline coordinate data to obtain the arc center offset. The molten pool contour image data is processed by edge detection and contour region centroid extraction algorithm to extract the center position of the molten pool contour. The position difference is then calculated with the welding trajectory centerline coordinate data to obtain the molten pool center offset. The fluctuation amplitude of welding current data, welding voltage data, wire feed speed data and welding speed data is calculated separately within the current welding window, and then the working condition disturbance is obtained by fusing the fluctuation statistics of sliding window and the time series correlation. The square of the difference in sidewall spacing plus one is used to obtain the sidewall constraint amplification. Calculate the welding torch centering deviation and add one to get the centering smoothing amount; calculate the sidewall constraint expansion amount, divide the result by the centering smoothing amount, add one, and take the natural logarithm to get the sidewall constraint logarithmic term. The absolute value of the difference between the arc center offset and the molten pool center offset is calculated to obtain the thermal input response deviation. The thermal input response deviation is calculated by dividing the result by the centering smoothing amount, and then the exponential function value is taken to obtain the thermal input deviation amplification term. The calculated operating condition disturbance is incremented by one to obtain the operating condition disturbance adjustment term; The arc offset constraint value is obtained by calculating the product of the logarithmic term of the sidewall constraint, the thermal input deviation amplification term, and the operating condition disturbance adjustment term.

5. The adaptive control method for welding torch posture based on ultra-thick-walled pipe welding according to claim 1, characterized in that: The specific process for determining the bias state and outputting the compensation level is as follows: Real-time comparison of arc deviation constraint values ​​and arc deviation constraint thresholds, which include primary constraint thresholds and secondary constraint thresholds: When the arc deviation constraint value is less than the secondary constraint threshold, maintain the current execution state corresponding to the welding torch end center coordinate data, welding torch forward tilt angle data, welding torch backward drag angle data and welding torch yaw angle data, and enter the equivalent heat input center reconstruction analysis. When the arc deviation constraint value is greater than or equal to the secondary constraint threshold and less than the primary constraint threshold, the deviation compensation mark is output. The coordinate data of the welding trajectory center line corresponding to the current welding section is used as the control reference and is not updated. The current output state corresponding to the welding current data and welding voltage data remains unchanged. The side wall spacing difference, arc center offset, molten pool center offset and welding torch centering deviation are sent to the equivalent heat input center reconstruction analysis. When the arc deviation constraint value is greater than or equal to the first-level constraint threshold, an enhanced compensation flag is output, limiting the swing amplitude of the welding torch yaw angle to the amplitude corresponding to the current enhanced compensation trigger moment. The current process state corresponding to the welding current data, welding voltage data, wire feed speed data, and welding speed data is frozen, and the side wall spacing difference, arc center offset, molten pool center offset, welding torch centering deviation, and operating condition disturbance are sent to the equivalent heat input center reconstruction analysis.

6. The adaptive control method for welding torch posture based on ultra-thick-walled pipe welding according to claim 1, characterized in that: The specific process of performing equivalent heat input center reconstruction analysis on the original observation data and prior control record data is as follows: The attitude unfolding of the welding torch tip center coordinate data, welding torch forward tilt angle data, welding torch backward drag angle data, welding torch yaw angle data, and welding wire tip spatial coordinate data is performed to obtain the incident position of the welding wire tip relative to the welding trajectory centerline coordinate data; the surface contour point data of the previous layer of weld bead are unfolded to obtain the surface incident reference corresponding to the current welding section; the historical archive execution results that match the current welding torch circumferential position angle data in the previous control record data are filtered to obtain the historical compensation reference results; the difference between the incident position and the surface incident reference is calculated to obtain the incident position deviation; the difference between the arc center offset and the molten pool center offset is calculated to obtain the thermal input response deviation. The difference between the sidewall spacing difference and the welding torch centering deviation is calculated to obtain the sidewall constraint mismatch; the difference between the candidate adjustment results corresponding to the current welding torch end center coordinate data, welding torch forward tilt angle data, welding torch backward drag angle data and welding torch yaw angle data and the historical compensation reference results is calculated to obtain the historical compensation deviation. The consistency of the direction of change and the consistency of the reduction magnitude of the incident position deviation, thermal input response deviation, sidewall constraint mismatch and historical compensation deviation are judged. The number of consistent items and the reduction magnitude are comprehensively mapped to obtain the reliable value of thermal input center reconstruction.

7. The adaptive control method for welding torch posture based on ultra-thick-walled pipe welding according to claim 1, characterized in that: The specific process of correcting the welding torch posture and regressing the heat input region based on the equivalent heat input center reconstruction analysis results is as follows: Real-time comparison of the confidence value and the confidence threshold for hot input center reconstruction: When the confidence value of the heat input center reconstruction is less than the confidence threshold of the heat input center reconstruction, the welding trajectory centerline coordinate data is used as the constraint reference to perform reverse deviation correction on the welding torch yaw angle data; the welding torch end center coordinate data is used as the control object to perform lateral micro-displacement correction on the welding torch end center coordinate data; the welding wire end spatial coordinate data and the previous layer weld surface contour point data are used as the basis to perform linkage correction on the welding torch forward tilt angle data and welding torch backward drag angle data; and the archived execution results of the current welding torch circumferential position angle data in the previous control record data are limited and called, and the corrected welding torch attitude correction results are sent to the convergence control analysis. When the confidence value of the heat input center reconstruction is greater than or equal to the confidence threshold of the heat input center reconstruction, a convergence reconstruction flag is output. The candidate adjustment results are kept from being updated with direction switching. The center coordinate data of the welding torch end is updated with a single position adjustment. The candidate adjustment results are sent to the convergence control analysis as the current welding torch attitude correction results.

8. The adaptive control method for welding torch posture based on ultra-thick-walled pipe welding according to claim 1, characterized in that: The specific process of executing control and performing convergence control analysis based on the welding torch attitude correction results is as follows: Based on the welding torch posture correction results, control is performed on the welding torch end center coordinate data, welding torch forward tilt angle data, welding torch backward drag angle data, and welding torch yaw angle data. After control is executed, the arc center position is re-extracted from the arc spot image data, and the molten pool center position is re-extracted from the molten pool contour image data. The re-extracted arc center position and the re-extracted molten pool center position are aligned with the welding trajectory centerline coordinate data to obtain the compensated position offset state. Then, the compensated position offset state within n consecutive welding windows is continuously processed to obtain the regression continuity discrimination result. The execution results that match the current layer number data and welding torch circumferential position angle data are extracted from the previous control record data to obtain the historical rollback reference results; The position difference between the compensated position offset state and the coordinate data of the welding trajectory centerline is calculated to obtain the offset change. Then, combined with the regression continuity judgment results and the historical backtracking reference results, the consistency of the regression direction and the reduction magnitude of the offset change is compared. The convergence control value is obtained by using the number of consistent terms and the reduction magnitude of the offset change as inputs.

9. The adaptive control method for welding torch posture based on ultra-thick-walled pipe welding according to claim 1, characterized in that: The specific process of performing closed-loop feedback control and weld retention control based on the convergence control analysis results is as follows: Real-time comparison of convergence control values ​​and convergence control thresholds: When the convergence control value is less than the convergence control threshold, the current output state corresponding to the welding current data and welding voltage data remains unchanged, the angle swing corresponding to the welding torch sway angle data is restricted from continuing to change, and the current welding torch posture correction result is sent back to the equivalent heat input center for reconstruction analysis. The welding torch posture correction result is regenerated in the interval corresponding to the welding torch circumferential position angle data with the same index granularity as the previous control record database. When the convergence control value is greater than or equal to the convergence control threshold, output the welding hold mark, maintain the current welding torch posture correction result and the current process execution status, and archive the current layer number data, welding torch circumferential position angle data, current welding torch posture correction result, and corresponding welding current data, welding voltage data, wire feed speed data and welding speed data to the preceding control record database.

10. An adaptive control system for welding torch posture based on ultra-thick-walled pipe welding, comprising the adaptive control method for welding torch posture based on ultra-thick-walled pipe welding as described in any one of claims 1-9, characterized in that, include: The acquisition and preprocessing module is used to acquire raw observation data during the operation of the welding equipment, obtain preceding control record data, and preprocess the raw observation data and preceding control record data. The asymmetric sidewall constraint discrimination module is used to jointly identify the spatial constraint differences on both sides of the bevel, the arc center offset characteristics, and the molten pool response offset characteristics based on the original observation data, and to perform edge state discrimination and compensation level output. The equivalent heat input center reconstruction module is used to perform equivalent heat input center reconstruction analysis on the original observation data and the preceding control record data, and to perform welding torch posture correction and heat input area regression control based on the equivalent heat input center reconstruction analysis results. The closed-loop verification and stability control module is used to perform control based on the welding torch attitude correction results and to conduct convergence control analysis. Based on the convergence control analysis results, it performs closed-loop feedback control and welding hold control.

Images