Intelligent welding control method and system for large transformer sheet structure

Abstract

The application discloses a large transformer sheet structure intelligent welding control method and system, and belongs to the technical field of welding control. The method comprises the following steps: constructing a welding thermal-mechanical coupling digital twin based on structured monitoring data, and performing real-time mapping on the welding process based on the welding thermal-mechanical coupling digital twin to obtain temperature field deviation information and deformation field deviation information; determining quality risk information corresponding to the current to-be-welded weld section based on the temperature field deviation information and the deformation field deviation information; based on the quality risk information, the welding parameters and the welding path are cooperatively optimized to obtain welding cooperative control information; and based on the welding cooperative control information, welding control is performed on the current weld section, and corresponding real-time welding feedback data is synchronously acquired.

Description

Technical Field

[0001] This application relates to the field of welding control technology, and more specifically, to an intelligent welding control method and system for a large transformer plate structure. Background Technology

[0002] Large transformers typically employ multi-pass welding to assemble sheet-like structural components such as tank walls, internal partitions, and heat sinks. This process involves long weld lengths and extended welding cycles. Furthermore, the low stiffness of these sheet-like components makes them susceptible to significant thermal deformation and residual stress due to welding heat input, which negatively impacts welding quality and the overall structural performance of the transformer. Current welding processes for large transformer sheet-like structures rely heavily on welding technicians' experience to pre-set welding parameters and paths. This approach makes it difficult to continuously and quantitatively monitor the dynamic changes in the temperature and deformation fields during welding. It also hinders timely adjustments to welding parameters and paths based on quality risks encountered during welding. Consequently, issues such as poor weld quality consistency, excessive post-weld deformation, and high rates of rework due to welding defects are prevalent.

[0003] In existing technologies, research has attempted to control the processing through digital twin technology. For example, patent application CN120145704A discloses a method, system, equipment, and storage medium for heat processing control based on digital twins. This patent application uses a multi-sensor fusion IoT data acquisition system to collect process parameters and quality data for smelting, casting, forging, and heat treatment processes, constructs a multi-dimensional process data platform for heat processing, and realizes parameter mapping and mesh transformation of multi-stage simulation results through an adaptive interface algorithm. Then, a digital twin model is constructed based on hierarchical parametric modeling, deployed to an edge computing gateway, and combined with a multi-round distributed process optimization engine to calculate the target process parameter combination. However, the solution disclosed in this patent application mainly focuses on the integration and collaborative simulation of process data for the entire hot processing process, including smelting, casting, forging, and heat treatment. It does not involve the real-time acquisition of temperature and deformation field deviations at the weld section level for the welding process itself, nor does it disclose specific technical means for collaboratively optimizing welding parameters and welding paths based on these deviations and for closed-loop dynamic adjustment of welding collaborative control information. Therefore, it is difficult to meet the needs of refined control of weld seams in the welding process of large transformer plate structures.

[0004] For example, patent application CN120124316A discloses a method and system for controlling the forging process based on digital twins. This patent application collects temperature and stress fields at multiple points in a partitioned manner for large hydraulic presses and forgings during the forging process. Based on the obtained multi-dimensional forging process database, it constructs a digital twin model of forging using a constitutive element nesting method. The forging material is divided into multiple macroscopic deformation units and embedded with microstructure evolution units, realizing bidirectional coupled simulation of macroscopic deformation and microstructure evolution. Then, the process parameter combination is determined by reverse solving through the process window. However, the scheme disclosed in this patent application mainly focuses on predicting the macroscopic and microscopic deformation and microstructure evolution of materials during the forging process and optimizing process parameters. It belongs to an offline optimization method for solving critical deformation conditions. It does not disclose the online correction of the digital twin based on real-time welding feedback data during the welding process, or the closed-loop control mechanism that continuously iterates and updates the modeling parameters and collaborative control rules of the digital twin by combining the data of the entire welding process and the post-weld quality inspection data. It also does not involve the real-time mapping and comparison of the deviation between the welding temperature field and the deformation field.

[0005] In addition, patent application CN122264443A discloses a method for scheduling and controlling the welding process of transformer radiators. This patent application constructs a digital twin disturbance model of welding heat input and generates a heat input disturbance factor matrix, thereby establishing an adaptive rearrangement mechanism for welding sequence under thermal field coupling constraints. By collecting micro-vibration and acoustic emission signals during the welding process, a hidden feature vector of welding quality is constructed to realize real-time judgment and trend prediction of welding quality. Furthermore, a global stress superposition heat map is generated through a cross-station stress coupling model to execute multi-station cross-equipment collaborative skip welding scheduling. Although this patent application also targets transformer welding scenarios, its technical solution focuses on the arrangement of welding process sequences and stress balance scheduling at multiple workstations and multiple equipment levels. It does not disclose the technical means of constructing a welding thermo-mechanical coupling digital twin for a single weld section and performing point-by-point real-time mapping and comparison of the temperature field and deformation field of that section. It also does not disclose the specific implementation method of multi-objective collaborative optimization of welding parameters and welding paths based on quality risk information, dynamic adjustment of collaborative control information combined with real-time welding feedback data, and online correction of the digital twin. Therefore, it is difficult to achieve continuous, quantitative monitoring and refined closed-loop control of temperature field and deformation field deviations at the weld level of large transformer plate structure.

[0006] In summary, the existing technology lacks an intelligent welding control method that can acquire the temperature field and deformation field deviation of the weld section in real time during the welding process of large transformer plate structures, and accordingly optimize the welding parameters and welding path through collaborative optimization and dynamic closed-loop control. At the same time, it supports online correction of digital twins and continuous iterative updates of modeling parameters and collaborative control rules. As a result, the problems of poor welding quality consistency, excessive deformation after welding, and high rework rate of welding defects in large transformer plate structures still exist. Summary of the Invention

[0007] In order to overcome a series of defects in the existing technology, the purpose of this application is to provide an intelligent welding control method for large transformer plate structures, which includes the following steps: A welding thermo-mechanical coupling digital twin is constructed based on structured monitoring data, and the welding process is mapped in real time based on the welding thermo-mechanical coupling digital twin to obtain temperature field deviation information and deformation field deviation information. Based on temperature field deviation information and deformation field deviation information, determine the quality risk information corresponding to the current weld section to be welded; Based on quality risk information, welding parameters and welding paths are collaboratively optimized to obtain welding collaborative control information. Based on welding collaborative control information, welding control is performed on the current weld section, and corresponding real-time welding feedback data is acquired simultaneously.

[0008] In some embodiments, the method for constructing a welding thermo-mechanical coupled digital twin based on structured monitoring data is as follows: Based on structured monitoring data and combined with historical structured monitoring data of similar welds, a welding thermo-mechanical coupling simulation model is constructed. The welding process was simulated and calculated using a welding thermo-mechanical coupling simulation model to obtain simulation data of temperature field and stress-strain field. A welding process state characterization model was constructed based on simulation data of temperature field and stress-strain field. By associating and configuring the welding thermo-mechanical coupling simulation model with the welding process state characterization model, a welding thermo-mechanical coupling digital twin is constructed. The welding thermo-mechanical coupling digital twin is driven and updated using real-time acquired structured monitoring data to characterize the real-time thermo-mechanical state of the welding process.

[0009] In some embodiments, the method for obtaining temperature field deviation information and deformation field deviation information is as follows: Welding process parameters are extracted from structured monitoring data according to a preset rolling time window; By inputting welding process parameters into the welding thermo-mechanical coupling digital twin, the temperature field prediction data and deformation field prediction data for the corresponding time window are obtained. Obtain the measured data corresponding to the temperature field prediction data and the deformation field prediction data; Determine temperature field deviation information and deformation field deviation information based on predicted data and measured data; The temperature field deviation information and deformation field deviation information are dynamically updated based on the welding process parameters updated according to the time window.

[0010] In some embodiments, the method for determining the quality risk information corresponding to the current weld section to be welded is as follows: Based on temperature field deviation information and deformation field deviation information, the risk indicators corresponding to the current weld section to be welded are determined. Based on risk indicators, determine the risk level corresponding to the current weld section to be welded; The dominant deviation type corresponding to the current weld section to be welded is determined based on the risk contribution item with the largest proportion in the risk indicators. Obtain the weld segment location information corresponding to the current weld segment to be welded; By associating the risk level, dominant deviation type, and weld section location information, quality risk information corresponding to the current weld section to be welded is generated.

[0011] In some embodiments, the method for collaboratively optimizing welding parameters and welding paths is as follows: Based on quality risk information, establish a collaborative optimization relationship between welding parameters and welding path; Based on the collaborative optimization relationship, multiple candidate collaborative optimization solutions are generated; The collaborative optimization scheme is evaluated and the target collaborative optimization scheme is determined. Based on the target collaborative optimization scheme, determine the welding parameter adjustment information and welding path adjustment information; Welding collaborative control information is generated based on welding parameter adjustment information and welding path adjustment information.

[0012] In some embodiments, the method for performing welding control on the current weld section is as follows: Welding control commands are generated based on welding collaborative control information; Adjust welding parameters and welding path according to welding control instructions; Control the welding actuator to move according to the adjusted welding path and maintain the preset relative posture relationship between the welding torch and the current weld section; Welding operations are performed on the current weld section based on the adjusted welding parameters and welding path.

[0013] In some embodiments, the intelligent welding control method further includes: Based on real-time welding feedback data, determine whether the current weld section meets the preset welding quality standard; If the target is not met, the welding collaborative control information will be dynamically adjusted based on real-time welding feedback data, and welding control will continue to be executed until the welding quality standard is met. If the target is reached, the welding of the current weld section is completed, and the welding thermo-mechanical coupling digital twin is corrected online based on real-time welding feedback data to obtain the corrected welding thermo-mechanical coupling digital twin.

[0014] In some embodiments, the method for dynamically adjusting welding collaborative control information is as follows: Based on real-time welding feedback data and target feedback data, welding feedback deviation information is determined; Based on welding feedback deviation information, determine the collaborative control adjustment amount; Based on the adjustment amount of the collaborative control, the current welding collaborative control information is corrected to obtain the adjusted welding collaborative control information; Based on the adjusted welding collaborative control information, corresponding welding control commands are generated. Welding control instructions are continuously updated based on real-time welding feedback data until the current weld section reaches the preset welding quality standard.

[0015] In some embodiments, the intelligent welding control method further includes: Based on the corrected welding thermo-mechanical coupling digital twin, and combined with the welding process data and post-weld quality inspection data, the modeling parameters and welding collaborative control rules of the welding thermo-mechanical coupling digital twin are iteratively updated to obtain the updated welding thermo-mechanical coupling digital twin and welding collaborative control rules.

[0016] The purpose of this application is also to provide an intelligent welding control system for a large transformer plate structure, used to implement the above-mentioned intelligent welding control method, including: The data acquisition and preprocessing module is used to acquire multi-source monitoring data during the welding process of large transformer plate structures, and to perform spatiotemporal registration and preprocessing on the multi-source monitoring data to obtain structured monitoring data. The digital twin construction module is communicatively connected to the data acquisition and preprocessing module. It is used to construct a welding thermo-mechanical coupling digital twin based on the structured monitoring data, and to perform real-time mapping of the welding process based on the welding thermo-mechanical coupling digital twin to obtain temperature field deviation information and deformation field deviation information. The quality risk assessment module is communicatively connected to the digital twin construction module and is used to determine the quality risk information corresponding to the current weld section to be welded based on the temperature field deviation information and the deformation field deviation information. The collaborative optimization module is communicatively connected to the quality risk assessment module and is used to collaboratively optimize welding parameters and welding paths based on the quality risk information to obtain welding collaborative control information. The welding control module is communicatively connected to the collaborative optimization module and is used to perform welding control on the current weld section based on the welding collaborative control information, and synchronously acquire the corresponding real-time welding feedback data.

[0017] Compared with existing technologies, this application has the following advantages: By constructing a welding thermo-mechanical coupling digital twin and mapping the welding process in real time, it is possible to continuously and quantitatively acquire temperature field and deformation field deviation information during the welding process, and determine the quality risk information corresponding to the current weld section accordingly; by collaboratively optimizing welding parameters and welding paths based on quality risk information, and dynamically adjusting welding collaborative control information based on real-time welding feedback data, closed-loop fine control of the welding process is achieved; on this basis, the digital twin and collaborative control rules are continuously iterated and updated by combining the welding process data and post-weld quality inspection data, thereby significantly improving the quality stability, welding qualification rate, and intelligent level of welding control in the welding process of large transformer plate structures. Attached Figure Description

[0018] Figure 1 A schematic diagram of the overall process of an intelligent welding control method for a large transformer plate structure provided in this application embodiment; Figure 2 A schematic diagram illustrating the arrangement of multi-source monitoring data acquisition devices and the process of acquiring structured monitoring data provided in the embodiments of this application; Figure 3 A schematic diagram illustrating the principle of constructing a welding thermo-mechanical coupling digital twin and real-time mapping of the welding process provided in the embodiments of this application; Figure 4 A schematic diagram illustrating the process for determining quality risk information and co-optimizing welding parameters and welding paths, as provided in the embodiments of this application; Figure 5 This is a schematic diagram illustrating the process of welding control execution, dynamic adjustment, online correction, and iterative update provided in the embodiments of this application. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of the embodiments of this invention will be described in more detail below with reference to the accompanying drawings. In the drawings, the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The described embodiments are some embodiments of this invention, but not all embodiments.

[0020] Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] The embodiments and directional terms described below with reference to the accompanying drawings are exemplary and intended to explain the present invention, and should not be construed as limiting the present invention.

[0022] Example 1 like Figure 1 As shown, this embodiment provides an intelligent welding control method for large transformer sheet structures, applied to the welding process control of sheet structures such as tank walls, internal partitions, and heat sinks in large transformers. The method includes the following steps S101 to S109: S101: During the welding process of large transformer plate structure, multi-source monitoring data is acquired, and the multi-source monitoring data is spatiotemporally registered and preprocessed to obtain structured monitoring data.

[0023] Specifically, the multi-source monitoring data includes at least temperature field distribution data, strain and displacement deformation data, visual image data of the weld pool, and welding current and voltage waveform data. These are collected by infrared thermal imagers mounted above or to the side of the welding area, strain sensors and non-contact displacement sensors attached to key monitoring points of the sheet structure, high-speed industrial cameras installed at the welding execution end, and current and voltage sampling modules built into the welding power supply. The multi-source monitoring data is time-registered based on the time series of the motion trajectory of the welding execution end, and then mapped to a unified three-dimensional coordinate system of the sheet structure to be welded to achieve spatial registration. After noise reduction filtering, outlier removal, and missing value interpolation, the data is resampled according to a preset spatiotemporal grid to obtain structured monitoring data with a unified spatiotemporal reference.

[0024] S102: Construct a welding thermo-mechanical coupling digital twin based on the structured monitoring data, and perform real-time mapping of the welding process based on the welding thermo-mechanical coupling digital twin to obtain temperature field deviation information and deformation field deviation information.

[0025] Specifically, based on the geometric parameters of the sheet structure design, material thermophysical parameters, and mechanical performance parameters contained in the structured monitoring data, a finite element thermo-mechanical coupling simulation model is established in conjunction with the structured monitoring data of similar historical welding. The simulation data of temperature field and stress-strain field are obtained by solving the model. Based on this, a proxy model reflecting the mapping relationship between welding process parameters, welding path, and temperature field and deformation field is established. The proxy model is driven in real time by the process parameters such as welding current, voltage, and speed collected in the structured monitoring data, so that it can characterize the real-time thermodynamic state of the welding process. According to the preset rolling time window, the process parameters within the time window are input into the digital twin to obtain the simulation prediction data of temperature field and deformation field. The simulation prediction data of temperature field and deformation field are compared point by point with the temperature field and deformation field data measured by infrared thermal imager and strain displacement sensor within the same time window. The residuals are used as temperature field deviation information and deformation field deviation information.

[0026] S103: Based on the temperature field deviation information and the deformation field deviation information, determine the quality risk information corresponding to the current weld section to be welded.

[0027] Specifically, a grading threshold is set for each deviation quantity in the temperature field deviation information and deformation field deviation information. The sub-risk levels corresponding to each deviation quantity are compared and converted into quantitative scores. The scores are then weighted and fused according to preset weight coefficients to obtain a comprehensive risk score, and the risk level (including low risk, medium risk, and high risk) corresponding to the current weld section is determined accordingly. At the same time, after normalizing each sub-deviation quantity, their numerical values ​​are compared, and the deviation category corresponding to the largest value is determined as the dominant deviation type. Combining the weld section location information such as the section number, start and end coordinates, and location type corresponding to the current weld section, the risk level, dominant deviation type, and weld section location information are correlated to generate the quality risk information corresponding to the current weld section to be welded.

[0028] S104: Based on the quality risk information, the welding parameters and welding path are optimized collaboratively to obtain welding collaborative control information.

[0029] Specifically, with the optimization objectives of reducing the current weld section quality risk score, controlling the total welding heat input, and ensuring welding efficiency, a multi-objective optimization model is constructed, which includes welding parameter adjustment and welding path adjustment. A multi-objective evolutionary algorithm is used to solve for several candidate collaborative optimization schemes distributed on the Pareto front. The optimal scheme is selected from the candidate schemes by combining the current weld section risk level and dominant deviation type, thus obtaining welding collaborative control information that includes parameter control sub-information and path control sub-information. Among them, the welding parameters include at least two of the following: welding current, welding voltage, welding speed, wire feed speed, welding torch oscillation amplitude, and interpass temperature.

[0030] S105: Based on the welding collaborative control information, perform welding control on the current weld section and obtain the corresponding real-time welding feedback data.

[0031] Specifically, the parameter control sub-information in the welding collaborative control information is sent to the welding power source and wire feeding mechanism, which then execute the welding action according to the sent welding current, voltage, and wire feeding speed; the path control sub-information is sent to the welding robot and the positioner linked with it, driving the welding torch to move along the adjusted welding path and maintain the preset relative welding posture; during the welding process, the characteristic data of arc voltage and current waveforms, the characteristic data of molten pool image, and the characteristic data of acoustic signals in the welding area are collected simultaneously as real-time welding feedback data.

[0032] S106: Based on the real-time welding feedback data, determine whether the current weld section meets the preset welding quality standard.

[0033] Specifically, the depth and width of the weld section are estimated based on the feature data of the molten pool image and compared with the allowable deviation range. The probability of porosity defects is assessed by combining acoustic signal feature data and compared with the upper limit of the probability. Based on the updated deformation field deviation information, it is determined whether it has converged to within the deformation field deviation convergence threshold. The results of the above three aspects are combined to determine whether the current weld section meets the preset welding quality standard.

[0034] S107: If the welding quality standard is not met, the welding collaborative control information is dynamically adjusted based on the real-time welding feedback data, and welding control continues to be executed until the welding quality standard is met.

[0035] Specifically, the deviation is obtained by comparing the real-time welding feedback data with the expected target feedback data of the welding collaborative control information at the corresponding time. Based on the preset welding process dynamic response model, the model predictive control method is used to solve the adjustment amount in a rolling manner with the goal of minimizing the weighted cost function of the deviation amount and parameter adjustment range within the preset future time window. The adjustment amount is then superimposed on the current welding collaborative control information and welding control continues to be executed. The above process is repeated according to the preset control cycle until the preset welding quality standard is met.

[0036] S108: If the welding quality standard has been met, the welding of the current weld section is completed, and the welding thermo-mechanical coupling digital twin is corrected online based on the real-time welding feedback data to obtain the corrected welding thermo-mechanical coupling digital twin.

[0037] Specifically, several modeling parameters that affect the simulation accuracy of the temperature field and deformation field in the digital twin are used as state variables to be corrected. An observation equation is established between the state variables and the real-time welding feedback data. The real-time welding feedback data obtained after the current weld section is completed is used as the observation. The state variables to be corrected are updated and estimated by Kalman filtering or Bayesian estimation method in combination with the state transition equation. The updated state variable values ​​are substituted into the digital twin to obtain the corrected welding thermo-mechanical coupling digital twin.

[0038] S109: Based on the corrected welding thermo-mechanical coupling digital twin, and combined with the welding process data and post-weld quality inspection data, the modeling parameters and welding collaborative control rules of the welding thermo-mechanical coupling digital twin are iteratively updated to obtain the updated welding thermo-mechanical coupling digital twin and welding collaborative control rules.

[0039] Specifically, the welding process data and post-weld quality inspection data corresponding to this welding task are added to the historical welding case library. Based on the updated historical welding case library, the network parameters of the neural network surrogate model in the digital twin are incrementally trained. The correlation between the parameters and path adjustment schemes corresponding to different dominant deviation types in the historical case library and the post-weld quality inspection data is statistically analyzed. The mapping rule library and the weight coefficients of the multi-objective optimization model used in the collaborative optimization step are corrected, thereby obtaining the updated welding thermo-mechanical coupling digital twin and welding collaborative control rules.

[0040] The intelligent welding control method for large transformer plate structures provided in this embodiment constructs a welding thermo-mechanical coupling digital twin and maps the welding process in real time. This allows for timely and accurate acquisition of temperature and deformation field deviations and quality risk information corresponding to the current weld section. By collaboratively optimizing welding parameters and welding paths based on quality risk information, and dynamically adjusting welding collaborative control information and online correcting the digital twin using real-time welding feedback data, closed-loop fine control of the welding process is achieved. Furthermore, by continuously iterating and updating the modeling parameters and collaborative control rules based on the entire welding process data and post-weld quality inspection data, the method significantly improves the quality consistency, welding pass rate, and intelligence and adaptability of the welding control system in the welding process of large transformer plate structures, meeting the actual needs of high-quality and high-efficiency welding control in large transformer manufacturing.

[0041] Example 2 like Figure 2 As shown, this embodiment, based on embodiment one, provides a more detailed explanation of the acquisition of multi-source monitoring data and the method of obtaining structured monitoring data in step S101.

[0042] Regarding the acquisition of multi-source monitoring data, key monitoring locations were identified for the welding process of large transformer sheet structures: Infrared thermal imagers, non-contactly mounted above or to the side of the sheet structure to be welded and maintaining a preset relative position with the welding execution end, were used to acquire temperature field distribution images of the welding area and its adjacent areas; strain sensors attached to key monitoring points on the sheet structure, and non-contact displacement sensors mounted near these key monitoring points, were used to acquire strain and displacement deformation data of the sheet body during welding; a high-speed industrial camera installed at the welding execution end was used to acquire visual images of the molten pool morphology, surface fluctuations, and spatter status; and the current and voltage sampling module built into the welding power supply was used to acquire real-time waveform data of the welding current and voltage. The aforementioned multi-source data, including temperature, strain-displacement, visual, and electrical signals, were synchronously acquired according to their respective inherent sampling frequencies and marked with a unified timestamp to provide a data foundation for subsequent spatiotemporal registration.

[0043] In terms of acquiring structured monitoring data, the motion trajectory time series of the welding execution end is used as the benchmark. Interpolation alignment is used to synchronously calibrate the monitoring data of different sampling frequencies such as temperature, deformation, visual, and electrical signals in the time dimension to eliminate the time delay difference between multi-source data. A unified three-dimensional coordinate system is established for the plate structure to be welded area. The spatial positions corresponding to the data collected by various monitoring devices such as infrared thermal imagers, strain sensors, and industrial cameras are mapped to this unified coordinate system to achieve spatial dimension registration of multi-source data. The registered data is then subjected to noise filtering, outlier removal, and missing value interpolation. The processed data is then resampled according to a preset spatiotemporal grid to finally obtain structured monitoring data with a unified time and spatial benchmark.

[0044] The intelligent welding control method for large transformer plate structures provided in this embodiment deploys infrared thermal imagers, strain and displacement sensors, high-speed industrial cameras, and current and voltage sampling modules for different monitoring objects, and timestamps the collected multi-source monitoring data to ensure comprehensive and synchronous acquisition of key status information during the welding process. Through spatiotemporal registration based on the welding execution end's motion trajectory, and preprocessing such as denoising, elimination, and interpolation, the time delay differences and spatial position deviations between multi-source data are eliminated, improving the accuracy and consistency of structured monitoring data. Based on this, high-quality data input is provided for the subsequent construction and real-time mapping of the welding thermo-mechanical coupling digital twin, thereby significantly improving the integrity and reliability of the monitoring data during the welding process of large transformer plate structures and the level of precision in welding quality control.

[0045] Example 3 like Figure 3As shown, this embodiment, based on embodiment one, elaborates on the construction method of the welding thermo-mechanical coupling digital twin in step S102 and the method of real-time mapping of the welding process and obtaining temperature field and deformation field deviation information.

[0046] In terms of digital twin construction, based on the geometric parameters, material thermophysical parameters, and mechanical performance parameters of the large transformer plate structure design contained in the structured monitoring data, and combined with historical similar welding structured monitoring data, a finite element thermo-mechanical coupling simulation model reflecting the movement law of welding heat source is established. The simulation datasets of temperature field and stress-strain field under different welding conditions are obtained by solving the model. Based on this simulation dataset, a neural network surrogate model is trained to learn the mapping relationship between welding process parameters, welding path, and temperature field and deformation field. This surrogate model replaces the computationally time-consuming solution link in the finite element model. The welding current, voltage, speed, and other process parameters collected in real time from the structured monitoring data are used as the input boundary conditions of the surrogate model for real-time driving. This enables the constructed digital twin to meet the computational efficiency requirements of real-time mapping of the welding process while ensuring a certain level of simulation accuracy.

[0047] Regarding real-time mapping and deviation information acquisition, according to a preset rolling time window, process parameters such as welding current, voltage, and speed from the structured monitoring data within the current time window are input into the welding thermo-mechanical coupling digital twin to obtain the simulation prediction results of the temperature field and deformation field corresponding to the time window. The simulation prediction results are then compared point by point with the temperature field and deformation field data synchronously measured by an infrared thermal imager and strain displacement sensor in the structured monitoring data. The residuals between the two at the corresponding spatial positions and time nodes are calculated, and these residuals are used as temperature field deviation information and deformation field deviation information. As the welding process progresses, the time window is updated by sliding at a preset step size, and the above mapping and comparison process is executed cyclically to achieve continuous dynamic tracking of welding process deviations.

[0048] The temperature field deviation information includes at least one of the following: peak temperature deviation, cooling rate deviation, and heat-affected zone width deviation. The peak temperature deviation is obtained by comparing the simulated peak temperature at the weld centerline with the measured peak temperature from an infrared thermal imager. The cooling rate deviation is obtained by comparing the slope of the simulated cooling curve within a preset temperature range with the slope of the measured cooling curve. The heat-affected zone width deviation is obtained by extracting contours from both the simulated and measured temperature fields according to preset critical temperature thresholds and comparing the widths of the extracted contours. The deformation field deviation information includes at least one of the following: transverse shrinkage deviation of the weld segment, angular deformation deviation, and overall flexural deformation deviation of the sheet structure. These are obtained by comparing the simulated predicted values ​​with the measured values ​​from strain-displacement sensors. All of the above deviations are statistically analyzed segmentally along the weld length, forming a deviation curve distributed along the weld direction to support the refined determination of subsequent quality risk information.

[0049] The intelligent welding control method for large transformer plate structures provided in this embodiment constructs a welding thermo-mechanical coupled digital twin by combining a finite element thermo-mechanical coupling simulation model and a neural network surrogate model. The surrogate model is driven by real-time process parameters, ensuring simulation accuracy while meeting the computational efficiency requirements of real-time mapping of the welding process. By comparing simulation prediction data with measured data point-by-point according to a rolling time window and extracting refined deviations such as peak temperature deviation, cooling rate deviation, heat-affected zone width deviation, lateral shrinkage deviation, angular deformation deviation, and overall flexural deformation deviation, continuous, quantitative, and dynamic tracking of temperature and deformation field deviations during the welding process is achieved. Based on this, accurate and reliable data support is provided for the refined judgment of quality risk information, thereby significantly improving the accuracy of deviation monitoring in the welding process of large transformer plate structures and the timeliness of welding quality risk early warning.

[0050] Example 4 like Figure 4 As shown, this embodiment, based on embodiment one, provides a detailed explanation of the method for determining quality risk information in step S103, the collaborative optimization of welding parameters and welding paths in step S104, and the method for executing welding control in step S105.

[0051] In determining quality risk information, corresponding grading thresholds are set for each deviation in the temperature field deviation information and deformation field deviation information. Each deviation is compared with the set grading thresholds to obtain the sub-risk level corresponding to each deviation. Each sub-risk level is then converted into a corresponding quantitative score according to a preset mapping relationship. The converted quantitative scores are weighted and fused according to preset weighting coefficients to obtain the comprehensive risk score corresponding to the current weld section. The comprehensive risk score is matched with a preset level range to determine the risk level corresponding to the current weld section. The risk level includes at least three levels: low risk, medium risk, and high risk. Different risk levels correspond to different subsequent collaborative optimization strategy selection methods and welding parameter adjustment range restrictions. Simultaneously, each sub-deviation quantity in the temperature field deviation information and deformation field deviation information is normalized, and the numerical values ​​of each normalized sub-deviation quantity are compared. The deviation category corresponding to the sub-deviation quantity with the largest value is determined as the dominant deviation type. The dominant deviation types include local overheating deviation, non-fusion tendency deviation, angular deformation deviation, joint shrinkage deviation, and overall deflection deviation. For different dominant deviation types, a mapping rule library of corresponding welding parameters and welding path adjustment directions is pre-established. In addition, the weld seam to be welded in the large transformer plate structure is pre-divided into several weld seam segments along its extension direction. Each segment is assigned a unique segment number and start and end coordinate values, and its cumulative welding length relative to the welding start point and its corner, end, or middle position type in the overall weld seam layout are recorded as weld seam segment location information. The risk level, dominant deviation type, and weld seam segment location information are stored in association to form a risk map distributed along the weld seam direction.

[0052] In terms of collaborative optimization, multiple optimization objectives are set, including reducing the current weld section quality risk score, controlling the total welding heat input, and ensuring welding efficiency. A multi-objective optimization model is constructed, incorporating welding parameter adjustments and welding path adjustments. A multi-objective evolutionary algorithm is used to solve this model, yielding a set of candidate collaborative optimization schemes distributed on the Pareto front. The optimal scheme is selected from the candidate schemes based on the current weld section's risk level and dominant deviation type, serving as the welding collaborative control information. The welding parameter adjustments and welding path adjustments are coupled variables and iterated together during the solution process. Specifically, when the dominant deviation type is excessive heat input, the welding current and voltage are reduced, the welding speed is increased, and the welding torch oscillation amplitude is reduced and the interpass waiting time is extended as needed to lower the interpass temperature. When the dominant deviation type is a lack of fusion tendency, the welding current is increased, the welding speed is appropriately reduced, and the welding torch oscillation amplitude is increased and the interpass temperature is appropriately increased as needed. While adjusting the above welding parameters, upper and lower limits corresponding to the risk level are set for the adjustment amplitude of each parameter. Regarding welding path optimization, when the dominant deviation type is overall deflection type deviation, the welding sequence of the subsequent weld sections to be welded is adjusted, and the welding path is rearranged by symmetrical segmentation and skip welding from the middle to both ends; when the dominant deviation type is angular deformation type deviation, the oscillation trajectory of the welding torch in the weld cross-section and the welding direction angle are adjusted to change the heat input distribution ratio of the molten pool on both sides of the weld plate.

[0053] In terms of welding control execution, the parameter control sub-information in the welding collaborative control information is sent to the welding power source and wire feeding mechanism. The welding power source and wire feeding mechanism execute the welding action according to the sent welding current, voltage and wire feeding speed instructions. The path control sub-information in the welding collaborative control information is sent to the welding robot and the positioner linked with it. The welding robot drives the welding torch to move along the current weld section according to the sent welding path and welding torch posture instructions. The positioner synchronously adjusts the posture and welding position of the plate structure. The welding robot and the positioner achieve linkage and coordination of trajectory and posture through a preset motion collaborative control strategy.

[0054] The intelligent welding control method for the large transformer plate structure provided in this embodiment achieves accurate identification of the quality risk level and the cause of the dominant deviation in the current weld section by classifying, quantifying, and weighting the various sub-deviations of the temperature field and deformation field, and determining the dominant deviation type by combining normalization comparison. By constructing a multi-objective collaborative optimization model aimed at reducing risk scores, controlling heat input, and ensuring welding efficiency, and by adopting differentiated parameter and path adjustment strategies for different dominant deviation types, the method achieves linkage and collaborative optimization of welding parameters and welding paths. On this basis, the collaborative optimization results are distributed to the welding power source, wire feeding mechanism, welding robot, and positioner to achieve trajectory and posture linkage, thereby significantly improving the pertinence of welding parameter and path adjustment, collaborative optimization efficiency, and the accuracy and response speed of welding control execution.

[0055] Example 5 like Figure 5 As shown, this embodiment, based on Embodiment 1, provides a detailed explanation of the method for judging welding quality standards in step S106, the method for dynamically adjusting welding collaborative control information in step S107, the method for online correction of digital twins in step S108, and the method for iterative updating of modeling parameters and collaborative control rules in step S109.

[0056] In terms of real-time welding feedback data acquisition, arc voltage and current waveform feature data are obtained by short-time window feature extraction of the original voltage and current signals acquired by the welding power source. The extracted features include waveform stability index and short-circuit transition frequency. Molten pool image feature data are obtained by image segmentation and edge extraction processing of molten pool images acquired by high-speed industrial cameras, including features such as molten pool length, width, and molten pool surface fluctuation amplitude. Acoustic signal feature data are obtained by time-frequency domain analysis of signals acquired by acoustic sensors set in the welding area, which is used to help determine the probability of welding defects such as spatter and porosity.

[0057] Regarding the judgment of welding quality standards, the preset welding quality standards include at least one of the following: allowable deviation range of weld penetration, allowable deviation range of weld width, upper limit of probability of porosity defects on weld surface, and convergence threshold of deformation field deviation corresponding to the current weld segment. Based on the molten pool image feature data in the real-time welding feedback data, the weld penetration and weld width of the current weld segment are estimated and compared with the corresponding allowable deviation range to obtain the dimensional quality judgment result. Combining acoustic signal feature data and molten pool image feature data, the probability of porosity defects occurring in the current weld segment is assessed and compared with the upper limit of the probability to obtain the defect risk judgment result. Based on the updated deformation field deviation information, it is determined whether it has converged to within the deformation field deviation convergence threshold to obtain the deformation control judgment result. If any of the above three judgment results does not meet the corresponding standard, it is determined that the current weld segment has not met the preset welding quality standard. The assessment result of the probability of porosity defects can be further verified and corrected in the post-weld quality inspection stage in combination with non-destructive testing data.

[0058] Regarding the dynamic adjustment of welding collaborative control information, real-time welding feedback data is compared with the target feedback data expected to be achieved by the welding collaborative control information at the corresponding time, and the deviation between the two is calculated. Based on the preset welding process dynamic response model, with the optimization objective of minimizing the weighted cost function of the deviation and the adjustment range of welding parameters within the preset future time window, the adjustment amount of welding collaborative control information is obtained by rolling solution using the model predictive control method. This adjustment amount is then superimposed on the current welding collaborative control information to obtain the adjusted welding collaborative control information, and welding control continues to be executed. The above process is executed cyclically according to the preset control cycle until the real-time welding feedback data meets the target range corresponding to the preset welding quality standard.

[0059] In terms of online calibration of digital twins, several modeling parameters affecting the simulation accuracy of temperature and deformation fields in the welding thermo-mechanical coupling digital twin are used as state variables to be calibrated. An observation equation reflecting the mapping relationship between the state variables to be calibrated and real-time welding feedback data is established. The real-time welding feedback data obtained after the current weld section is welded is used as the observation. Combined with the state transition equation corresponding to the state variable, Kalman filtering or Bayesian estimation methods are used to update and estimate the state variables to be calibrated. The updated estimated state variable values ​​are substituted into the welding thermo-mechanical coupling digital twin to obtain the calibrated welding thermo-mechanical coupling digital twin, thereby improving its simulation prediction accuracy for subsequent weld sections to be welded.

[0060] Regarding the construction and iterative updating of the welding process database, the welding process data includes multi-source monitoring data, structured monitoring data, quality risk information, welding collaborative control information, and real-time welding feedback data corresponding to each weld section. After being indexed according to time sequence and weld section number, a data set is formed. Post-weld quality inspection data is obtained by using at least one non-destructive testing method, such as ultrasonic testing or radiographic testing, to obtain the location, type, and size information of internal defects in the weld of the large transformer sheet structure after welding. The location, type, and size information of surface opening defects in the weld are obtained by using penetrant testing. The welding process data and post-weld quality inspection data are associated and matched according to the weld section number to establish a database of correspondence between welding process parameters and paths and the final internal and surface quality of the weld. Based on this, the welding process data and post-weld quality inspection data corresponding to this welding task were added to the historical welding case library. The network parameters of the neural network surrogate model in the digital twin were incrementally trained using an online learning method to improve its simulation prediction generalization ability under different welding conditions. At the same time, the correlation between the welding parameters and path adjustment schemes corresponding to different dominant deviation types in the historical welding case library and the final post-weld quality inspection data were statistically analyzed. The mapping rule library and the weight coefficients of the multi-objective optimization model used in the collaborative optimization step were corrected to obtain the updated welding collaborative control rules.

[0061] Furthermore, the large transformer plate structure includes at least one of the following plate steel structural components: transformer tank wall panels, internal partitions, and heat sinks. The types of weld joints involved include at least one of the following: butt joints, T-joints, and corner joints. For different plate structural components and different joint types, the modeling parameters of the welding thermo-mechanical coupling digital twin and the welding heat source model adopt appropriate parameter values. The grading threshold in the quality risk information and the weight coefficient of the multi-objective optimization model in the collaborative optimization are also pre-set with differentiated initial values ​​for different plate structural components and joint types to adapt to the actual situation that the welding thermo-mechanical response characteristics are different due to the differences in structural stiffness and heat dissipation conditions in different parts of the large transformer plate structure.

[0062] The intelligent welding control method for large transformer plate structures provided in this embodiment improves the accuracy and comprehensiveness of online welding quality assessment by extracting multi-dimensional feature data such as arc voltage and current waveforms, molten pool images, and acoustic signals, and combining them with preset quality standards to comprehensively judge weld size, defect risk, and deformation convergence. It achieves dual closed-loop optimization of welding process control and digital twin simulation accuracy by dynamically adjusting welding collaborative control information based on model predictive control methods and performing online correction of the digital twin based on Kalman filtering or Bayesian estimation methods. Furthermore, it iteratively updates modeling parameters and collaborative control rules by continuously accumulating the welding process database, and sets differentiated parameters for different plate structure components and joint types. This significantly improves the adaptability, long-term operational reliability, and welding quality control accuracy of the intelligent welding control method for large transformer plate structures under different welding conditions, meeting the actual needs of high-quality and high-efficiency welding control for diverse plate structure components in the manufacturing process of large transformers.

[0063] Example 6 To verify the effectiveness of the intelligent welding control method for large transformer plate structures provided in this application under actual working conditions, the following explanation uses the welding of the tank wall panel of a 500MVA power transformer from a transformer manufacturing company as an example.

[0064] The transformer tank wall panel is constructed from Q345R steel plates, with the weld section numbered WB-2-07 (weld section location information: start and end coordinates (1200, 350) to (1200, 2200), location type being a butt weld in the middle of the wall panel). The plate thickness is 28mm, the weld length is 1850mm, and the joint type is a butt joint. During the welding process of this weld section in September 2024, a welding thermo-mechanical coupling digital twin was constructed based on the technical solution of this application, and the entire welding process was controlled to achieve comprehensive perception and dynamic control of welding quality risks.

[0065] First, using an infrared thermal imager (model FLIR A700) mounted above the weld section, strain gauges attached to key monitoring points (a total of 12 monitoring points), and a non-contact laser displacement sensor, combined with a high-speed industrial camera (sampling frame rate 200fps) installed at the welding execution end and the built-in current and voltage sampling module of the welding power supply (model Lincoln Power Wave S700), the temperature field distribution data, strain and displacement deformation data, molten pool visual image data, and current and voltage waveform data of the weld section during the welding process are collected synchronously. After spatiotemporal registration and preprocessing, structured monitoring data is obtained.

[0066] During the construction and real-time mapping of the digital twin, a finite element thermo-mechanical coupling simulation model (approximately 86,000 mesh nodes) was established based on the geometric parameters, thermophysical parameters, and mechanical properties of the weld section, combined with historical structural monitoring data of similar welds. A neural network surrogate model was then trained based on this model (approximately 12,000 training samples). Real-time process parameters were input into the digital twin at preset rolling time windows (window length 2 seconds) to obtain simulation predictions of the temperature and deformation fields. These predictions were then compared point-by-point with synchronously measured data from an infrared thermal imager and strain displacement sensors. The calculations showed that the peak temperature deviation of the weld section at the 620mm mark was 38℃, the cooling rate deviation was 12%, the transverse shrinkage deviation was 0.42mm, and the angular deformation deviation was 0.31°.

[0067] Based on the aforementioned deviation information, the comparison, quantification, and weighted fusion were performed according to preset grading thresholds, resulting in a comprehensive risk score of 76 points for the weld section, which was matched and determined to be at the "medium risk" level. Simultaneously, normalization comparison determined the dominant deviation type to be "angular deformation type deviation." After generating quality risk information by combining the location information of the weld section, collaborative optimization of welding parameters and welding path was triggered: through multi-objective evolutionary algorithm solving and combined with risk level and dominant deviation type screening, the final determined welding collaborative control information was to reduce the welding current from the original set value of 215A to 198A, increase the welding speed from 350mm / min to 380mm / min, and adjust the welding sequence of subsequent welded sections from the original sequential welding to symmetrical segmentation, with skip welding from the middle to both ends.

[0068] After the above welding collaborative control information is sent to the welding power source and welding robot for execution, real-time welding feedback data is continuously collected. The detection shows that the weld depth of this section is 21.3 mm and the weld width is 9.8 mm, both of which are within the preset allowable deviation range. The probability assessment value of porosity defects is 1.2%, which is lower than the upper limit of the probability of occurrence of 2%. The deformation field deviation has converged to within the preset convergence threshold. It is determined that the weld section has reached the preset welding quality standard, and the welding of the current weld section is completed.

[0069] After welding was completed, based on real-time welding feedback data for that section, the Kalman filter method was used to perform online correction of the modeling parameters in the welding thermo-mechanical coupling digital twin. After correction, the temperature field prediction error of the digital twin for subsequent adjacent weld sections was reduced from 9.6% to 4.1%. Simultaneously, the data from the entire welding process, along with post-weld quality inspection data obtained from ultrasonic testing and penetrant testing, were added to the historical welding case library. Based on this, incremental training of the neural network surrogate model was performed, and the weight coefficients of the collaborative optimization mapping rule base and the multi-objective optimization model were corrected.

[0070] Statistics show that after adopting the technical solution of this application, the first-pass yield rate of 42 weld sections in the same batch of transformer tank panels increased from 91.3% to 98.6% compared to similar products before the adoption. The average overall deflection deformation after welding decreased from 1.86 mm to 0.74 mm, the welding defect repair rate decreased from 7.5% to 1.4%, and the average welding control response time of a single weld section was less than 1.5 seconds. Compared with the traditional welding method that does not adopt the technical solution of this application and relies solely on manual experience to set welding parameters and paths, the intelligent welding control method for large transformer plate structures provided by this application shows significant advantages in terms of welding quality consistency, post-weld deformation control accuracy, and welding defect repair rate, verifying the effectiveness and engineering applicability of this method in the actual welding production of large transformer plate structures.

[0071] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A smart welding control method for a large transformer plate structure, characterized in that, Includes the following steps: A welding thermo-mechanical coupling digital twin is constructed based on structured monitoring data, and the welding process is mapped in real time based on the welding thermo-mechanical coupling digital twin to obtain temperature field deviation information and deformation field deviation information. Based on temperature field deviation information and deformation field deviation information, determine the quality risk information corresponding to the current weld section to be welded; Based on quality risk information, welding parameters and welding paths are collaboratively optimized to obtain welding collaborative control information. Based on welding collaborative control information, welding control is performed on the current weld section, and corresponding real-time welding feedback data is acquired simultaneously.

2. The intelligent welding control method according to claim 1, characterized in that, The method for constructing a welding thermo-mechanical coupled digital twin based on structured monitoring data is as follows: Based on structured monitoring data and combined with historical structured monitoring data of similar welds, a welding thermo-mechanical coupling simulation model is constructed. The welding process was simulated and calculated using a welding thermo-mechanical coupling simulation model to obtain simulation data of temperature field and stress-strain field. A welding process state characterization model was constructed based on simulation data of temperature field and stress-strain field. By associating and configuring the welding thermo-mechanical coupling simulation model with the welding process state characterization model, a welding thermo-mechanical coupling digital twin is constructed. The welding thermo-mechanical coupling digital twin is driven and updated using real-time acquired structured monitoring data to characterize the real-time thermo-mechanical state of the welding process.

3. The intelligent welding control method according to claim 1, characterized in that, The methods for obtaining temperature field deviation information and deformation field deviation information are as follows: Welding process parameters are extracted from structured monitoring data according to a preset rolling time window; By inputting welding process parameters into the welding thermo-mechanical coupling digital twin, the temperature field prediction data and deformation field prediction data for the corresponding time window are obtained. Obtain the measured data corresponding to the temperature field prediction data and the deformation field prediction data; Determine temperature field deviation information and deformation field deviation information based on predicted data and measured data; The temperature field deviation information and deformation field deviation information are dynamically updated based on the welding process parameters updated according to the time window.

4. The intelligent welding control method according to claim 1, characterized in that, The method for determining the quality risk information corresponding to the current weld section to be welded is as follows: Based on temperature field deviation information and deformation field deviation information, the risk indicators corresponding to the current weld section to be welded are determined. Based on risk indicators, determine the risk level corresponding to the current weld section to be welded; The dominant deviation type corresponding to the current weld section to be welded is determined based on the risk contribution item with the largest proportion in the risk indicators. Obtain the weld segment location information corresponding to the current weld segment to be welded; By associating the risk level, dominant deviation type, and weld section location information, quality risk information corresponding to the current weld section to be welded is generated.

5. The intelligent welding control method according to claim 1, characterized in that, The method for collaboratively optimizing welding parameters and welding paths is as follows: Based on quality risk information, establish a collaborative optimization relationship between welding parameters and welding path; Based on the collaborative optimization relationship, multiple candidate collaborative optimization solutions are generated; The collaborative optimization scheme is evaluated and the target collaborative optimization scheme is determined. Based on the target collaborative optimization scheme, determine the welding parameter adjustment information and welding path adjustment information; Welding collaborative control information is generated based on welding parameter adjustment information and welding path adjustment information.

6. The intelligent welding control method according to claim 1, characterized in that, The method for implementing welding control in the current weld section is as follows: Welding control commands are generated based on welding collaborative control information; Adjust welding parameters and welding path according to welding control instructions; Control the welding actuator to move according to the adjusted welding path and maintain the preset relative posture relationship between the welding torch and the current weld section; Welding operations are performed on the current weld section based on the adjusted welding parameters and welding path.

7. The intelligent welding control method according to any one of claims 1-6, characterized in that, The intelligent welding control method further includes: Based on real-time welding feedback data, determine whether the current weld section meets the preset welding quality standard; If the target is not met, the welding collaborative control information will be dynamically adjusted based on real-time welding feedback data, and welding control will continue to be executed until the welding quality standard is met. If the target is reached, the welding of the current weld section is completed, and the welding thermo-mechanical coupling digital twin is corrected online based on real-time welding feedback data to obtain the corrected welding thermo-mechanical coupling digital twin.

8. The intelligent welding control method according to claim 7, characterized in that, The method for dynamically adjusting welding collaborative control information is as follows: Based on real-time welding feedback data and target feedback data, welding feedback deviation information is determined; Based on welding feedback deviation information, determine the collaborative control adjustment amount; Based on the adjustment amount of the collaborative control, the current welding collaborative control information is corrected to obtain the adjusted welding collaborative control information; Based on the adjusted welding collaborative control information, corresponding welding control commands are generated. Welding control instructions are continuously updated based on real-time welding feedback data until the current weld section reaches the preset welding quality standard.

9. The intelligent welding control method according to claim 8, characterized in that, The intelligent welding control method further includes: Based on the corrected welding thermo-mechanical coupling digital twin, and combined with the welding process data and post-weld quality inspection data, the modeling parameters and welding collaborative control rules of the welding thermo-mechanical coupling digital twin are iteratively updated to obtain the updated welding thermo-mechanical coupling digital twin and welding collaborative control rules.

10. An intelligent welding control system for a large transformer plate structure, used to implement the intelligent welding control method as described in claim 9, characterized in that, include: The data acquisition and preprocessing module is used to acquire multi-source monitoring data during the welding process of large transformer plate structures, and to perform spatiotemporal registration and preprocessing on the multi-source monitoring data to obtain structured monitoring data. The digital twin construction module is communicatively connected to the data acquisition and preprocessing module. It is used to construct a welding thermo-mechanical coupling digital twin based on the structured monitoring data, and to perform real-time mapping of the welding process based on the welding thermo-mechanical coupling digital twin to obtain temperature field deviation information and deformation field deviation information. The quality risk assessment module is communicatively connected to the digital twin construction module and is used to determine the quality risk information corresponding to the current weld section to be welded based on the temperature field deviation information and the deformation field deviation information. The collaborative optimization module is communicatively connected to the quality risk assessment module and is used to collaboratively optimize welding parameters and welding paths based on the quality risk information to obtain welding collaborative control information. The welding control module is communicatively connected to the collaborative optimization module and is used to perform welding control on the current weld section based on the welding collaborative control information, and synchronously acquire the corresponding real-time welding feedback data.

Images