Metal component welding stress dynamic regulation method and system based on digital twinning

Abstract

The application discloses a metal component welding stress dynamic regulation method and system based on digital twinning, relates to the intelligent control technical field of a welding process, and comprises the following steps: collecting original welding signals in a welding process in real time, and synchronously constructing a welding process digital twinning body interacting with the original welding signals; simulating and calculating the prototype stress field distribution of a metal component inside a current welding stage in the twinning body; generating welding energy dynamic regulation instructions according to comparison between the distribution and a preset stress threshold interval, and calculating instant welding energy correction parameters to be fed to a welding execution mechanism to adjust energy input; after adjustment, collecting updated signals again to drive the twinning body to perform simulation verification, obtaining a verified stress field distribution, and performing closed-loop calibration on the aforementioned correction parameters based on the verified stress field distribution, so that calibrated parameters are formed and used for dynamic regulation in subsequent stages. The application realizes real-time perception, prediction and active accurate control of welding internal stress.

Description

Technical Field

[0001] This invention belongs to the field of intelligent control technology for welding processes, specifically a method and system for dynamic control of welding stress in metal components based on digital twins. Background Technology

[0002] In the field of automated or semi-automated welding of metal components, controlling welding stress and deformation is crucial to ensuring component performance and dimensional accuracy. Existing technologies mainly employ two methods: relying on process experiments and experience to preset fixed process parameters before welding; and during the welding process, using sensors to monitor external measurable signals such as arc voltage, current, and molten pool images, and adjusting some process parameters based on the feedback values ​​of these signals to maintain apparent stability of the welding process.

[0003] Existing technologies have limitations. The root cause of welding stress and deformation lies in the dynamically changing temperature and stress fields within the component due to thermal cycling. These fields cannot be directly measured online using conventional sensors. Current online control methods can only perform closed-loop adjustments to externally measurable parameters. Their control logic is disconnected from the generation mechanism of internal stress states, making it an indirect and lagging control approach. Due to the lack of real-time perception and prediction capabilities of the internal stress evolution process, existing methods cannot intervene before the risk of stress exceeding limits actually occurs. They can only remedy the situation through subsequent processes after welding or passively accept potential deformations and defects.

[0004] The problem this invention aims to solve is to achieve direct insight and proactive intervention into the internal stress generation process during welding. This requires establishing a method capable of mapping and calculating the internal stress field in real time, and using this as the core basis to drive the dynamic adjustment of welding energy input. This allows for direct control of the stress evolution path during welding, realizing a shift from passively stabilizing the external process to actively shaping the internal state. Summary of the Invention

[0005] This invention aims to solve at least one of the technical problems existing in the prior art; Therefore, this invention proposes a method for dynamic control of welding stress in metal components based on digital twins, comprising: Real-time acquisition of raw welding signals during automatic or semi-automatic electric arc or plasma arc welding processes; A digital twin of the welding process is constructed synchronously based on the original welding signals and interacts with the physical welding process in real time. The prototype stress field distribution inside the metal component at the current welding stage is simulated and calculated in the digital twin of the welding process. Based on the comparison between the prototype stress field distribution and the preset stress threshold range, a dynamic control command for welding energy is generated. The instantaneous welding energy correction parameters are calculated based on the welding energy dynamic control command, and the instantaneous welding energy correction parameters are sent to the welding execution mechanism. The welding actuator adjusts the welding energy input based on the instantaneous welding energy correction parameters; After adjusting the welding energy input, the updated welding signal is collected again, and the digital twin of the welding process is driven to perform simulation verification to obtain the stress field distribution after verification. Based on the verified stress field distribution, the instantaneous welding energy correction parameters are calibrated in a closed loop to form calibrated welding energy parameters for dynamic control in subsequent welding stages.

[0006] Furthermore, the real-time acquisition of raw welding signals during the automatic or semi-automatic electric arc or plasma arc welding process includes: By deploying a multi-source sensor array on welding equipment and metal components, welding current signals, welding voltage signals, welding pool image sequences, infrared thermal images of component surfaces, and acoustic emission signals of key points of components are simultaneously acquired. The collected welding current and welding voltage signals are filtered and normalized to generate standard electrical signal waveforms. Edge features are extracted and morphological analysis is performed on the acquired weld pool image sequence to generate dynamic geometric parameters of the weld pool; The temperature field of the collected infrared thermal images of the component surface is reconstructed to generate welding thermal cycle curves; Time-frequency analysis was performed on the acoustic emission signals of key points of the components to extract acoustic emission features related to microstructure changes; The original welding signal is formed by integrating the standard electrical signal waveform, dynamic geometric parameters of the molten pool, welding thermal cycle curve, and acoustic emission characteristics. The process of reconstructing the temperature field from the acquired infrared thermal image of the component surface includes: The infrared thermal image is calibrated for pixel temperature, and the grayscale values ​​are converted into temperature values. By combining the spatial position parameters of the sensor, the two-dimensional temperature value is mapped onto the three-dimensional geometric model of the metal component surface; An interpolation algorithm is used to fill the measurement blind zone and generate a continuous and complete welding temperature field on the component surface.

[0007] Furthermore, the construction of a digital twin of the welding process based on the original welding signal, which interacts in real time with the physical welding process, includes: An initial twin framework was established, including a material property database, a welding thermophysical model, and structural constraints. The standard electrical signal waveform and welding thermal cycle curve in the original welding signal are input into the welding thermophysical model to calculate the spatial distribution and transient evolution of the welding heat input; The dynamic geometric parameters of the molten pool in the original welding signal are used as boundary conditions to drive the iterative calculation of the molten pool fluid dynamics model in the initial twin frame; The acoustic emission characteristics in the original welding signal are correlated with the phase transformation dynamic parameters in the material property database to correct the microstructure evolution model in the initial twin framework; By integrating the calculation results of the welding thermophysical model, the iterative results of the molten pool fluid dynamics model, and the modified microstructure evolution model, a digital twin of the welding process is formed that maintains data synchronization and interaction with the physical welding process.

[0008] Furthermore, the simulation and calculation of the prototype stress field distribution inside the metal component at the current welding stage in the digital twin of the welding process includes: In the digital twin of the welding process, the transient temperature field of the metal component is calculated by thermo-elastic-plastic finite element analysis based on the heat input distribution, molten pool morphology and microstructure prediction results of the current welding stage. Calculate the thermal strain field caused by non-uniform thermal expansion based on the transient temperature field; By combining the thermal strain field with the structural constraints and material constitutive relations in the digital twin of the welding process, the prototype stress field distribution inside the metal component under the action of welding thermal cycle is calculated by solving the mechanical equilibrium equation.

[0009] Furthermore, the step of generating a dynamic welding energy control command based on the comparison between the prototype stress field distribution and the preset stress threshold range includes: Extract the maximum tensile stress value and the stress gradient of the key region from the prototype stress field distribution; The maximum tensile stress value is compared with a preset allowable tensile stress threshold, and the stress gradient of the key region is compared with a preset allowable stress gradient threshold. If the maximum tensile stress value exceeds the allowable tensile stress threshold, an energy reduction control command is generated to reduce heat input. If the stress gradient in the critical region exceeds the allowable stress gradient threshold, an energy redistribution control command is generated to improve the uniformity of heat distribution. If neither the maximum tensile stress value nor the stress gradient of the critical region exceeds the corresponding threshold, an energy stability control command is generated to maintain the current parameters.

[0010] Furthermore, the instantaneous welding energy correction parameters calculated based on the welding energy dynamic control command include: When the energy reduction control command is received, the reduction amount of welding current or welding voltage is calculated by proportional integral algorithm based on the proportion by which the maximum tensile stress value exceeds the allowable tensile stress threshold, and used as an instant welding energy correction parameter. When the energy redistribution control command is received, the welding speed or welding torch oscillation amplitude is adjusted through a path planning algorithm based on the stress gradient distribution of the key area to generate instantaneous welding energy correction parameters for improving heat distribution. When the energy stabilization control command is received, the current welding parameters are set as the instant welding energy correction parameters.

[0011] Furthermore, after adjusting the welding energy input, the updated welding signal is acquired again, and the digital twin of the welding process is driven to perform simulation verification to obtain the verified stress field distribution, including: After performing welding energy adjustment based on the instantaneous welding energy correction parameters, and after a preset response delay time, the updated welding signal during the welding process is collected again by the multi-source sensor array. The updated welding signal is input into the welding process digital twin to update the heat input distribution, molten pool morphology, and microstructure in the welding process digital twin; In the updated digital twin of the welding process, a thermo-elastic-plastic finite element analysis was performed again to calculate the verified stress field distribution of the metal component after adjusting the welding energy input.

[0012] Furthermore, the closed-loop calibration of the instantaneous welding energy correction parameters based on the verified stress field distribution includes: Calculate the residual between the verified stress field distribution and the expected stress field distribution; The validity of the instantaneous welding energy correction parameter is determined based on the magnitude and sign of the residual. If the residual exceeds the allowable range, the instantaneous welding energy correction parameter is calculated in reverse using a feedback control algorithm based on the residual to generate the calibrated welding energy parameter. If the residual is within the allowable range, then the instantaneous welding energy correction parameter is confirmed to be the calibrated welding energy parameter.

[0013] Furthermore, by combining the thermal strain field with the structural constraints and material constitutive relations in the digital twin of the welding process, and solving the mechanical equilibrium equations, the prototype stress field distribution inside the metal component under the welding thermal cycle is calculated, including: The calculated thermal strain field is used as the initial strain load. The structural constraints formed by the fixture positioning points and rigid support points defined in the digital twin of the welding process are applied. The material constitutive relation is defined by the material yield strength, elastic modulus, and coefficient of thermal expansion as varying with temperature in the material property database. A set of mechanical equilibrium equations with nodal displacements as unknowns is established in the finite element solver, and geometric nonlinearity and material nonlinearity are considered for iterative solution. Finally, the prototype stress field distribution inside the metal component is output.

[0014] Furthermore, the present invention also includes a dynamic control system for welding stress of metal components based on digital twins. The system includes a memory, a processor, and a computer program stored in the memory and running on the processor. When the processor executes the computer program, it implements the steps of the dynamic control method for welding stress of metal components based on digital twins as described above.

[0015] Compared with the prior art, the beneficial effects of the present invention are: By synchronously driving the digital twin of the welding process based on real-time acquired welding signals, the prototype stress field distribution inside the metal component can be directly calculated. This technique transforms the internal mechanical state, which was previously impossible to measure directly during welding, into continuous and quantitative digital information. Traditional methods, unable to obtain this internal state, lack direct target guidance for their control behavior. Through real-time simulation of the twin, a dynamic data source of internal stress construction is provided for welding process control for the first time, enabling the control system's decision-making basis to delve into the intrinsic causes from external appearance parameters, achieving real-time quantitative perception of the root causes of welding stress generation.

[0016] Based on the comparison between the prototype stress field output by the digital twin and the preset threshold, dynamic control commands for welding energy are directly generated, realizing feedforward control based on stress prediction. This logic changes the passive response mode of traditional feedback control, allowing proactive intervention before the stress field deviates from the allowable range. The generation of control commands directly aims to suppress or regulate the stress evolution in a specific region, giving the adjustment of energy input clear physical meaning and specificity. This allows for more effective intervention in the stress accumulation process, exerting influence from the source.

[0017] After energy adjustment, the updated welding signal is immediately used to drive the digital twin for rapid simulation verification, and the control parameters are calibrated in a closed loop based on the verification results. This process constructs an online iterative loop of "decision-execution-evaluation-optimization". The effect of each adjustment action can be quickly and virtually evaluated, and its deviation can be used to correct the control model or parameter mapping relationship in real time. This enables the entire control system to continuously learn and adjust itself during the welding process, dynamically adapting to nonlinear factors such as changes in material properties and fluctuations in operating conditions, and gradually improving the stress control accuracy and adaptability for the specific component and process. Attached Figure Description

[0018] Figure 1 This is a flowchart illustrating the steps of the method for dynamic control of welding stress in metal components based on digital twins as described in this invention. Figure 2 A flowchart for the simulation calculation of the stress field distribution of the prototype; Figure 3 A flowchart for generating dynamic control commands for welding energy; Figure 4 Stress comparison curves controlled by PI algorithm for multi-layer welding of low-alloy high-strength steel thick plates; Figure 5 Stress field distribution diagram after verification during the aluminum alloy welding calibration stage. Detailed Implementation

[0019] The technical solution of the present invention will be clearly and completely described below with reference to the embodiments. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] See Figure 1The method acquires raw welding signals in real time during automatic or semi-automatic electric arc and plasma arc welding processes. These raw welding signals are synchronously used to construct a digital twin of the welding process that interacts with the physical welding process in real time. Within this digital twin, the prototype stress field distribution inside the metal component at the current welding stage is simulated and calculated. This prototype stress field distribution is compared with a preset stress threshold range, and a dynamic control command for welding energy is generated based on the comparison result. According to this control command, the instantaneous welding energy correction parameters are calculated and sent to the welding actuator. The welding actuator adjusts the actual welding energy input accordingly. After the energy adjustment is completed, the method acquires the updated welding signals again and drives the welding process digital twin to perform simulation verification to obtain the verified stress field distribution. Based on this verified stress field distribution, the previously issued instantaneous welding energy correction parameters are calibrated in a closed loop to form calibrated welding energy parameters, which are used to guide the dynamic control process in subsequent welding stages.

[0021] In one embodiment of the present invention, real-time acquisition of raw welding signals during automatic or semi-automatic electric arc or plasma arc welding is achieved through a multi-source sensor array arranged on the welding equipment and metal components. The multi-source sensor array simultaneously acquires welding current signals, welding voltage signals, weld pool image sequences, infrared thermograms of the component surface, and acoustic emission signals from key points of the component. The acquired welding current and welding voltage signals are filtered and normalized. The filtering process uses a low-pass filter to remove high-frequency noise, and the normalization process adjusts the signal amplitude to a standard range, generating a standard electrical signal waveform. Edge feature extraction and morphological analysis are performed on the acquired weld pool image sequences. Edge feature extraction uses the Canny operator to detect the weld pool boundary, and morphological analysis smooths the boundary and extracts geometric features through expansion and erosion operations, generating dynamic geometric parameters of the weld pool, including the weld pool length, width, and area. Temperature field reconstruction is performed on the acquired infrared thermograms of the component surface. This reconstruction includes pixel-level temperature calibration of the infrared thermograms, converting grayscale values ​​into temperature values, and mapping the two-dimensional temperature values ​​onto a three-dimensional geometric model of the metal component surface using the sensor's spatial position parameters. A bilinear interpolation algorithm is then used to fill measurement blind spots, generating a continuous and complete welding temperature field on the component surface and deriving the welding thermal cycle curve. Time-frequency analysis is performed on the acoustic emission signals from key points of the component. This analysis uses short-time Fourier transform to extract signal features in the time and frequency domains, extracting acoustic emission features related to microstructure changes, including acoustic emission energy count and peak frequency. The original welding signal is then formed by integrating standard electrical signal waveforms, molten pool dynamic geometric parameters, welding thermal cycle curves, and acoustic emission features.

[0022] In some embodiments, the formula for converting grayscale values ​​to temperature values ​​during temperature field reconstruction is expressed as follows: in: Indicates the temperature value. This represents the grayscale value of an infrared thermal image. Indicates the temperature calibration factor. This represents the temperature offset coefficient. It can be understood as the temperature calibration coefficient. and temperature offset coefficient It is determined in advance through a blackbody radiation calibration experiment.

[0023] A digital twin of the welding process is constructed synchronously with the original welding signal and interacts in real time with the physical welding process. In specific implementation, an initial twin framework is established, including a material property database, a welding thermophysical model, and structural constraints. The material property database stores the material thermophysical parameters of the metal components, such as thermal conductivity, specific heat capacity, and density. The standard electrical signal waveform and welding thermal cycle curve from the original welding signal are input into the welding thermophysical model, which calculates the spatial distribution and transient evolution of the welding heat input based on the heat conduction equation. The dynamic geometric parameters of the molten pool in the original welding signal are used as boundary conditions to drive the iterative calculation of the molten pool fluid dynamics model in the initial twin framework. The molten pool fluid dynamics model simulates the fluid flow and thermal convection effects within the molten pool. The acoustic emission characteristics in the original welding signal are correlated with the phase transformation dynamic parameters in the material property database to modify the microstructure evolution model in the initial twin framework. The microstructure evolution model predicts the phase transformation process and grain growth in the welding heat-affected zone. By integrating the calculation results of the welding thermophysical model, the iterative results of the molten pool fluid dynamics model, and the modified microstructure evolution model, a digital twin of the welding process is formed that maintains data synchronization and interaction with the physical welding process.

[0024] Optionally, the arrangement of the multi-source sensor array is optimized according to the welding process and component geometry, for example, symmetrically arranging infrared thermal imagers on both sides of the weld to obtain a full-view temperature field. It is understood that the frame rate of the weld pool image sequence is matched with the welding speed to ensure the temporal continuity of the dynamic geometric parameters of the weld pool. In some embodiments, filtering and normalization processing are performed in real time in the embedded system using digital signal processing algorithms to reduce data transmission latency. Optionally, the extraction of acoustic emission features also includes waveform parameter analysis such as rise time and duration to enhance the correlation with microstructural changes.

[0025] In practical implementation, data comparison is reflected in the processing of the original welding signal before and after processing. For example, the unfiltered welding current signal contains power supply noise and unstable arc fluctuations. After filtering and normalization, the standard electrical signal waveform exhibits smooth periodic characteristics, which is beneficial for subsequent model input. A comparison of edge feature extraction before and after processing the weld pool image sequence shows that the weld pool boundary in the original image is blurred and affected by spatter. After edge feature extraction and morphological analysis, the dynamic geometric parameters of the weld pool have clear contours and quantized values. A comparison of temperature field reconstruction before and after processing the infrared thermogram of the component surface shows that the original infrared thermogram only provides a two-dimensional grayscale distribution. After temperature calibration, three-dimensional mapping, and interpolation, a continuous and complete welding temperature field on the component surface is generated, which can reflect the spatial gradient of the welding thermal cycle.

[0026] See Figure 2 In one embodiment of the present invention, the prototype stress field distribution inside the metal component at the current welding stage is simulated and calculated in a digital twin of the welding process. Specifically, based on the predicted results of the heat input distribution, molten pool morphology, and microstructure at the current welding stage, the transient temperature field of the metal component is calculated using thermo-elastic-plastic finite element analysis (TEL). TEL discretizes the geometric model of the metal component in the digital twin of the welding process into a finite element mesh. The thermal strain field caused by non-uniform thermal expansion is calculated based on the transient temperature field. The calculation of the thermal strain field is based on the temperature change value at each node in the temperature field and the thermal expansion coefficient of the material. Combining the thermal strain field with the structural constraints and material constitutive relations in the digital twin of the welding process, the prototype stress field distribution inside the metal component is calculated by solving the mechanical equilibrium equations.

[0027] In practical implementation, combining the thermal strain field with the structural constraints and material constitutive relations in the digital twin of the welding process, the calculation of the prototype stress field distribution inside the metal component under the welding thermal cycle involves the following steps: The calculated thermal strain field is applied as the initial strain load to the corresponding nodes of the finite element model. Structural constraints formed by the fixture positioning points and rigid support points defined in the digital twin of the welding process are applied to the corresponding nodes of the finite element model in the form of fixed displacement boundary conditions. The material constitutive relations defined by the material yield strength, elastic modulus, and coefficient of thermal expansion varying with temperature are called from the material property database. In the finite element analysis, the material constitutive relations are expressed as temperature-dependent elastoplastic stress-strain curves. A set of mechanical equilibrium equations with nodal displacements as unknowns is established in the finite element solver, and geometric and material nonlinearities are considered for iterative solution, ultimately outputting the prototype stress field distribution inside the metal component.

[0028] In some embodiments, a specific example scenario is butt welding of low-carbon steel plates. In this scenario, the heat input distribution received by the digital twin of the welding process at the current welding stage is the parameter of the double ellipsoidal heat source model. The predicted molten pool morphology is the weld width and weld depth, and the predicted microstructure is the austenite grain size. Based on these inputs, thermo-elastic-plastic finite element analysis is performed. The calculated transient temperature field shows that the peak temperature at the center of the weld pool is 1720 degrees Celsius, and the temperature gradient of the heat-affected zone is 450 degrees Celsius / mm. Based on this transient temperature field, the thermal strain field is calculated, and the calculated thermal strain value near the fusion line reaches 0.008. Data comparison shows that the calculated peak temperature near the fusion line is compared with the temperature inverted from the synchronously acquired infrared thermogram. The calculated value is 1720 degrees Celsius, while the infrared measurement value is 1695 degrees Celsius, with an absolute deviation of 25 degrees Celsius. The calculated width of the heat-affected zone is compared with the actual width measured by the metallographic experiment. The calculated value is 5.2 mm, while the measured value is 5.0 mm.

[0029] In some embodiments, the iterative solution process of the mechanical equilibrium equations is expressed by the following formula: in: This represents the tangent stiffness matrix for the current load step. This represents the nodal displacement increment vector. Represents the external load vector. This represents the internal resistance vector. It can be understood as the tangent stiffness matrix. Composed of the current material Jacobian matrix and element geometry, internal resistance vector This is obtained by integrating the current stress state. In the example scenario, the prototype stress field distribution output after solving the above equation shows that the peak value of the longitudinal residual tensile stress in the weld is 350 MPa. This calculated value can be compared with subsequent X-ray diffraction measurements.

[0030] Optionally, the structural constraints can be defined as applying completely fixed constraints at both ends of the plate and displacement constraints perpendicular to the plate surface at several support points at the bottom of the plate. It can be understood that calling the material constitutive relation means that the stress and stiffness updates at each integration point are obtained by querying the material property database based on the current temperature at that point. In specific implementations, data comparison is further reflected in the difference between the prototype stress field distribution and the stress field distribution after subsequent verification. For example, in the same example scenario, the transverse stress distribution curve of the weld centerline obtained from the initial calculation and the corresponding curve recalculated by the digital twin after an energy adjustment show quantifiable differences in shape and peak value. This difference is used to drive closed-loop calibration.

[0031] Optionally, geometric nonlinearity considerations include updating the shape functions of the finite element to account for large displacement effects, while material nonlinearity considerations include using kinematic hardening criteria to describe the plastic behavior of the material. In practice, the prototype stress field distribution is output as a contour plot data file, containing the stress tensor components of all nodes or integration points in the model, for subsequent comparison with a preset stress threshold range.

[0032] See Figure 3 In one embodiment of the present invention, a dynamic control command for welding energy is generated based on the comparison between the prototype stress field distribution and a preset stress threshold range. The maximum tensile stress value and the stress gradient of key areas in the prototype stress field distribution calculated and output by the digital twin of the welding process are extracted. The key areas are pre-defined according to the welding process specifications or component design requirements, for example, a range of 10 mm on each side of the weld centerline. The extracted maximum tensile stress value is compared with a preset allowable tensile stress threshold, which is determined based on the yield strength of the base material and a safety factor. The extracted stress gradient of the key areas is compared with a preset allowable stress gradient threshold, which is set based on the requirements for preventing welding cracks and controlling deformation.

[0033] In practice, if the maximum tensile stress value in the prototype stress field distribution exceeds a preset allowable tensile stress threshold, an energy reduction control command is generated to reduce welding heat input. If the stress gradient in a critical region of the prototype stress field distribution exceeds a preset allowable stress gradient threshold, an energy redistribution control command is generated to improve the uniformity of heat distribution. If neither the maximum tensile stress value nor the stress gradient in a critical region of the prototype stress field distribution exceeds its corresponding preset threshold, an energy stability control command is generated to maintain the current welding parameters.

[0034] In some embodiments, a specific example scenario is the welding of fillet welds in low-alloy high-strength steel box girders. In this scenario, the calculated maximum tensile stress extracted from the prototype stress field distribution is 420 MPa, and the preset allowable tensile stress threshold is 350 MPa. Data comparison shows that the calculated maximum tensile stress exceeds the allowable tensile stress threshold by 70 MPa. The critical region is defined as the area adjacent to the weld toe. The calculated stress gradient in this region extracted from the prototype stress field distribution is 28 MPa / mm, and the preset allowable stress gradient threshold is 25 MPa / mm. Data comparison shows that the calculated stress gradient in the critical region exceeds the allowable stress gradient threshold by 3 MPa / mm. Based on the above comparison results, both the maximum tensile stress value and the stress gradient in the critical region exceed the corresponding thresholds. The control logic will simultaneously generate energy reduction control commands and energy redistribution control commands, and determine the priority of the commands or fuse them.

[0035] Optionally, the comparison between the maximum tensile stress value and the allowable tensile stress threshold can be expressed mathematically: if This triggers an energy reduction and regulation command, in which This represents the calculated maximum tensile stress value extracted from the prototype stress field distribution. This indicates the preset allowable tensile stress threshold. This is understandable. It is a scalar value derived from a traversal search of the first principal stress at all nodes in the prototype stress field distribution.

[0036] In some embodiments, another example scenario is the splicing welding of aluminum alloy thin plates. In this scenario, the calculated maximum tensile stress extracted from the prototype stress field distribution is 180 MPa, and the preset allowable tensile stress threshold is 200 MPa. Data comparison shows that the calculated maximum tensile stress is less than the allowable tensile stress threshold. The critical region is defined as the entire weld heat-affected zone, and the extracted stress gradient is calculated to be 15 MPa / mm², while the preset allowable stress gradient threshold is 20 MPa / mm². Data comparison shows that the calculated stress gradient in the critical region is less than the allowable stress gradient threshold. Based on the two comparison results, neither the maximum tensile stress value nor the stress gradient in the critical region exceeds the threshold, therefore, an energy stabilization control command is generated.

[0037] In practice, the stress gradient is calculated by sampling stress values ​​along a path perpendicular to the welding direction within a pre-defined key area, and using the central difference method to calculate the gradient at each sampling point. The gradient with the largest absolute value is taken as the stress gradient for comparison. Data comparison exists not only between calculated values ​​and thresholds, but also between different types of control commands generated at different welding stages. For example, an energy stabilization control command may be generated at the beginning of box girder welding, while an energy reduction control command is generated in the middle stage when welding heat accumulates. This switching of command types reflects the dynamic change of the prototype stress field distribution as the welding process progresses.

[0038] Optionally, the preset allowable stress gradient threshold may not be a fixed value, but rather multiple segmented thresholds set according to different locations of the component or different phase transformation temperature ranges of the material. It can be understood that the generation logic of the energy redistribution control command primarily focuses on the drastic degree of stress change in space, rather than the absolute level of stress. In specific implementation, the generated welding energy dynamic control command is a structured data object containing the command type, triggering reason (based on stress exceeding limits or gradient exceeding limits), and the corresponding original stress data. This command is transmitted to the subsequent calculation module to solve for the real-time welding energy correction parameters.

[0039] In one embodiment of the present invention, real-time welding energy correction parameters are calculated based on the welding energy dynamic control command. When an energy reduction control command is received, the reduction in welding current or welding voltage is calculated using a proportional-integral algorithm based on the proportion by which the maximum tensile stress value exceeds the allowable tensile stress threshold. The proportional-integral algorithm uses the current stress error and the error integral term for weighted calculation. When an energy redistribution control command is received, the welding speed or welding torch oscillation amplitude is adjusted using a path planning algorithm based on the stress gradient distribution in the critical area. The path planning algorithm replans the welding torch movement path or adjusts the oscillation mode according to the high-value areas of the stress gradient distribution. When an energy stabilization control command is received, the current welding parameters are set as real-time welding energy correction parameters. The current welding parameters include welding current, welding voltage, welding speed, and welding torch oscillation amplitude.

[0040] In some embodiments, a specific example scenario is multi-layer welding of thick plates made of low-alloy high-strength steel. In this scenario, the welding energy dynamic control command is an energy reduction control command, based on the fact that the maximum tensile stress value in the prototype stress field distribution calculated in the previous stage is 410 MPa, and the allowable tensile stress threshold is 350 MPa. in: This indicates the maximum tensile stress value. This indicates the allowable tensile stress threshold. The calculated value exceeds the allowable threshold. The reduction in welding current is calculated using the proportional-integral (PI) algorithm. The formula for the PI algorithm is as follows: ,in This indicates the amount of reduction in welding current. Represents the proportionality coefficient. Represents the integral coefficient. This indicates that the historical value exceeds the cumulative proportional value. Set the proportional coefficient. The integral coefficient is -15 amperes per unit. The reduction in welding current is calculated based on a ratio of -5 amperes per unit and a historical cumulative excess ratio of 0.05. Amperes. The data comparison is reflected in the welding current values ​​before and after adjustment. The welding current before adjustment was 280 amperes. Based on the calculated real-time welding energy correction parameters, the welding current was adjusted to approximately 277.2 amperes.

[0041] In practical implementation, the path planning algorithm adjusts the welding speed or welding torch oscillation amplitude in response to energy redistribution control commands. The algorithm reads the stress gradient distribution map of key areas and identifies regions where the stress gradient exceeds a local threshold. The algorithm generates new welding torch trajectories or modifies oscillation parameters to relatively reduce welding heat input in areas of high stress gradient and relatively increase it in areas of low stress gradient. For example, it increases the welding speed or reduces the oscillation dwell time in areas of high stress gradient, and decreases the welding speed or increases the oscillation amplitude in areas of low stress gradient. Real-time welding energy correction parameters are generated to improve heat distribution. These parameters include a set of welding speed or oscillation amplitude settings that vary with time or location.

[0042] In some embodiments, a specific example scenario is stainless steel pipe welding. In this scenario, the dynamic control command for welding energy is an energy redistribution control command, stemming from the uneven distribution of stress gradient in the critical area, with a local peak gradient reaching 30 MPa / mm, exceeding the allowable stress gradient threshold of 25 MPa / mm. The path planning algorithm, based on the stress gradient distribution map, divides the welding path into several segments and calculates a welding speed correction factor for each segment. Referring to Table 1, examples of some instantaneous welding energy correction parameters output by the path planning algorithm are shown.

[0043] Table 1: Example Table of Energy Redistribution Control Parameters for Stainless Steel Pipe Welding The data comparison is reflected before and after the welding speed adjustment. In the 50-100 mm and 150-200 mm sections with high stress gradients, the welding speed was increased to 138 mm / min and 144 mm / min, respectively, with the aim of reducing local heat input by passing through the high gradient area faster.

[0044] Optionally, the historical values ​​exceeding the proportional-to-integral (PII) in the proportional-to-integral (PII) algorithm can be used. Updated in each adjustment cycle, the update rule is to adjust the excess ratio of the current cycle. It is accumulated into historical values. This can be understood as a scaling factor. and integral coefficient The sign is negative because the welding energy input needs to be reduced when the stress exceeds the threshold. In practice, data comparison is also reflected in the switching of different control command types. For example, in the initial stage of thick plate welding, an energy stabilization control command may be generated, and the welding parameters remain unchanged; as heat accumulates, an energy reduction control command is generated, and the welding current is reduced; in the final stage of weld seam, in order to control deformation, an energy redistribution control command may be generated to adjust the welding speed.

[0045] Optionally, the welding torch oscillation amplitude correction parameters generated by the path planning algorithm can be expressed as adjustments to the oscillation frequency and oscillation width. For example, reducing the oscillation width in areas with high stress gradients can concentrate heat input. It can be understood that when both energy reduction control commands and energy redistribution control commands are received simultaneously, the calculation module will integrate the two commands. For example, it might first calculate the reduction in welding current based on the energy reduction control command, and then plan the speed or oscillation adjustment based on the adjusted current according to the energy redistribution control command. In practical implementation, the real-time welding energy correction parameters are sent to the welding actuator in the form of digital commands. The welding actuator parses the commands and adjusts the corresponding physical parameters, such as power output or servo motor movement.

[0046] See Figure 4 In the verification of stress dynamic control in multi-layer welding of thick low-alloy high-strength steel plates, the figure compares the evolution of maximum tensile stress during the welding process under two conditions: PI algorithm control and no PI control. Specifically, the red dashed curve without PI control shows that the maximum tensile stress continuously rises after welding begins, eventually exceeding the allowable tensile stress threshold (350 MPa, green dashed line) and climbing above 410 MPa, exhibiting a significant stress accumulation trend. In contrast, the blue solid curve with PI algorithm control, after the control takes effect, dynamically corrects the welding energy input through a proportional-integral algorithm, stabilizing the maximum tensile stress below 350 MPa, fluctuating only within the 340 MPa~350 MPa range, effectively avoiding the risk of stress exceeding limits. From a temporal perspective, after approximately 10 seconds, the PI control takes effect, and the stress difference between the two curves continues to widen, intuitively verifying the effectiveness of the proportional-integral algorithm in adjusting the welding current by real-time calculation of stress error and error integral term. This demonstrates the core role of the digital twin-based dynamic stress control system in suppressing the accumulation of welding tensile stress.

[0047] In one embodiment of the present invention, after adjusting the welding energy input, the updated welding signal is acquired again and the welding process digital twin is driven to perform simulation verification to obtain the verified stress field distribution. After performing welding energy adjustment based on real-time welding energy correction parameters, a preset response delay time is allowed to wait for the welding process to reach a new quasi-steady state. The updated welding signal during the welding process is acquired again through a multi-source sensor array. The updated welding signal includes the adjusted welding current signal, welding voltage signal, weld pool image sequence, infrared thermogram of the component surface, and acoustic emission signals of key points of the component. The updated welding signal is input into the welding process digital twin to update the heat input distribution, weld pool morphology, and microstructure in the welding process digital twin. The heat input distribution is updated according to the new standard electrical signal waveform and welding thermal cycle curve. In the updated welding process digital twin, thermo-elastic-plastic finite element analysis is performed again. The thermo-elastic-plastic finite element analysis uses the same model and boundary conditions as the calculation prototype stress field distribution to calculate the verified stress field distribution of the metal component after adjusting the welding energy input.

[0048] In practice, the instant welding energy correction parameters are calibrated in a closed loop based on the verified stress field distribution. The residual between the verified stress field distribution and the expected stress field distribution is calculated. The residual is the field distribution formed by the difference between the nodal stress values ​​of the verified stress field distribution and the corresponding nodal stress values ​​of the expected stress field distribution. The validity of the instant welding energy correction parameters is determined by the magnitude and sign of the residual. The magnitude of the residual is quantified by the root mean square error or the maximum absolute value error, and the sign of the residual indicates whether the stress deviation is higher or lower than the expected value. If the residual exceeds the allowable range, the instant welding energy correction parameters are calculated in reverse using a feedback control algorithm based on the residual to generate calibrated welding energy parameters. If the residual is within the allowable range, the instant welding energy correction parameters are confirmed as calibrated welding energy parameters.

[0049] In some embodiments, a specific example scenario is the welding of a thin-walled aluminum alloy structure. In this scenario, the preset response delay time is 3 seconds. After 3 seconds, a multi-source sensor array acquires a set of updated welding signals. Data comparison shows that the average value of the welding current signal acquired before adjustment was 252 amperes, while the average value of the updated welding current signal acquired after adjustment was 245 amperes, confirming that the welding actuator has implemented the instantaneous welding energy correction parameters to reduce the current. After inputting the updated welding signal into the welding process digital twin, the recalculated and verified stress field distribution shows that the maximum tensile stress at the weld center is 332 MPa. The desired stress field distribution in this scenario is set as a homogenized target stress field, with a maximum tensile stress expected to be 320 MPa at the weld center. The residual is calculated, and the stress residual at this location is +12 MPa. The root mean square error of the entire stress field residual is 8.7 MPa, while the preset allowable range is 5.0 MPa.

[0050] It is understandable that residuals exceeding the allowable range indicate that the instantaneous welding energy correction parameters failed to bring the stress field sufficiently close to the desired state. Based on the residuals, reverse compensation calculations are performed using a feedback control algorithm. This feedback control algorithm employs an incremental PID form, and its compensation calculation formula is as follows: in: This indicates the additional compensation amount for the welding current correction. This represents the root mean square error of the stress field residuals during the current calculation period. Indicates proportional gain. Indicates integral gain. Represents differential gain. Represents the integral term of error. This represents the differential term of the error. (Setting) -0.3 amperes per megapascal, -0.05 amperes per megapascal second, The value is 0, the current error integral term is 25 MPa second, and the additional compensation amount is calculated. Amperes. The final calibrated welding energy parameter is the instantaneous welding energy correction parameter (245 Amperes) plus an additional compensation (-3.86 Amperes), which is approximately 241.1 Amperes.

[0051] In some embodiments, another example scenario is the welding of low-carbon steel plates. In this scenario, the root mean square error of the residual between the verified stress field distribution and the desired stress field distribution is 4.2 MPa, which is less than the allowable range of 5.0 MPa. Data comparison shows that the residual is within the allowable range, therefore the instantaneous welding energy correction parameter is determined to be effective. This instantaneous welding energy correction parameter is confirmed as the calibrated welding energy parameter and is directly used for control in subsequent welding stages. Optionally, the desired stress field distribution can be a constant target value field or a theoretically optimal stress field dynamically generated according to the welding process specification. It can be understood that the integral term in the feedback control algorithm... Used to eliminate steady-state error, differential term It is used to predict the trend of error changes. In practice, the data comparison process is continuous. For example, in each adjustment cycle, a new post-verification stress field distribution is calculated and compared with the expected stress field distribution, so that a new round of calibration can be performed based on the parameters calibrated in the previous cycle.

[0052] See Figure 5 The results presented are the verification results of the stress field after closed-loop calibration of dynamic control of welding stress based on digital twins. Specifically, the stress field exhibits a symmetrical gradient distribution with the weld center (X≈10mm, Y≈0mm) as the core: the stress value is highest in the weld center region, reaching approximately 376 MPa, appearing as a red-orange cluster area; gradually transitioning to yellow, green, and cyan regions towards the periphery, the stress value decreases to 352 MPa, 336 MPa, and 328 MPa respectively; in the edge region far from the weld (X<5mm or X>15mm, Y<-3mm or Y>3mm), the stress value drops back to the blue base range of 304~320 MPa. This distribution characteristic reflects the spatial attenuation law of welding heat input, that is, the weld center generates significant thermal strain due to concentrated heat input, thus forming a high tensile stress zone, while the stress level decreases in the peripheral region due to the weakening of thermal influence. The color gradient and numerical distribution of the stress field in the figure can be directly used to verify the effectiveness of the instantaneous welding energy correction parameters: if this stress field is compared with the expected stress field, the root mean square error of the residual can quantify the calibration effect of the current parameters and provide data support for subsequent feedback compensation.

[0053] The above embodiments are only used to illustrate the technical methods of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical methods of the present invention without departing from the spirit and scope of the technical methods of the present invention.

Claims

1. A method for dynamic control of welding stress in metal components based on digital twins, characterized in that, The method includes: Real-time acquisition of raw welding signals during automatic or semi-automatic electric arc or plasma arc welding processes; A digital twin of the welding process is constructed synchronously based on the original welding signals and interacts with the physical welding process in real time. The simulation calculation of the prototype stress field distribution inside the metal component at the current welding stage in the digital twin of the welding process includes: In the digital twin of the welding process, the transient temperature field of the metal component is calculated by thermo-elastic-plastic finite element analysis based on the heat input distribution, molten pool morphology and microstructure prediction results of the current welding stage. Calculate the thermal strain field caused by non-uniform thermal expansion based on the transient temperature field; By combining the thermal strain field with the structural constraints and material constitutive relations in the digital twin of the welding process, the prototype stress field distribution inside the metal component under the action of welding thermal cycle is calculated by solving the mechanical equilibrium equation. Based on the comparison between the prototype stress field distribution and the preset stress threshold range, a dynamic control command for welding energy is generated, including: Extract the maximum tensile stress value and the stress gradient of the key region from the prototype stress field distribution; The maximum tensile stress value is compared with a preset allowable tensile stress threshold, and the stress gradient of the key region is compared with a preset allowable stress gradient threshold. If the maximum tensile stress value exceeds the allowable tensile stress threshold, an energy reduction control command is generated to reduce heat input. If the stress gradient in the critical region exceeds the allowable stress gradient threshold, an energy redistribution control command is generated to improve the uniformity of heat distribution. If neither the maximum tensile stress value nor the stress gradient of the critical region exceeds the corresponding threshold, an energy stability control command is generated to maintain the current parameters. The instantaneous welding energy correction parameters are calculated based on the welding energy dynamic control command, and the instantaneous welding energy correction parameters are sent to the welding execution mechanism. The welding actuator adjusts the welding energy input based on the instantaneous welding energy correction parameters; After adjusting the welding energy input, the updated welding signal is collected again, and the digital twin of the welding process is driven to perform simulation verification to obtain the stress field distribution after verification. Based on the verified stress field distribution, the instantaneous welding energy correction parameters are calibrated in a closed loop to form calibrated welding energy parameters for dynamic control in subsequent welding stages.

2. The method for dynamic control of welding stress in metal components based on digital twins as described in claim 1, characterized in that, The real-time acquisition of raw welding signals during automatic or semi-automatic electric arc or plasma arc welding processes includes: By deploying a multi-source sensor array on welding equipment and metal components, welding current signals, welding voltage signals, welding pool image sequences, infrared thermal images of component surfaces, and acoustic emission signals of key points of components are simultaneously acquired. The collected welding current and welding voltage signals are filtered and normalized to generate standard electrical signal waveforms. Edge features are extracted and morphological analysis is performed on the acquired weld pool image sequence to generate dynamic geometric parameters of the weld pool; The temperature field of the collected infrared thermal images of the component surface is reconstructed to generate welding thermal cycle curves; Time-frequency analysis was performed on the acoustic emission signals of key points of the components to extract acoustic emission features related to microstructure changes; The original welding signal is formed by integrating the standard electrical signal waveform, dynamic geometric parameters of the molten pool, welding thermal cycle curve, and acoustic emission characteristics. The process of reconstructing the temperature field from the acquired infrared thermal image of the component surface includes: The infrared thermal image is calibrated for pixel temperature, and the grayscale values ​​are converted into temperature values. By combining the spatial position parameters of the sensor, the two-dimensional temperature value is mapped onto the three-dimensional geometric model of the metal component surface; An interpolation algorithm is used to fill the measurement blind zone and generate a continuous and complete welding temperature field on the component surface.

3. The method for dynamic control of welding stress in metal components based on digital twins as described in claim 1, characterized in that, The digital twin of the welding process, which is synchronously constructed based on the original welding signal and interacts with the physical welding process in real time, includes: An initial twin framework was established, including a material property database, a welding thermophysical model, and structural constraints. The standard electrical signal waveform and welding thermal cycle curve in the original welding signal are input into the welding thermophysical model to calculate the spatial distribution and transient evolution of the welding heat input; The dynamic geometric parameters of the molten pool in the original welding signal are used as boundary conditions to drive the iterative calculation of the molten pool fluid dynamics model in the initial twin frame; The acoustic emission characteristics in the original welding signal are correlated with the phase transformation dynamic parameters in the material property database to correct the microstructure evolution model in the initial twin framework; By integrating the calculation results of the welding thermophysical model, the iterative results of the molten pool fluid dynamics model, and the modified microstructure evolution model, a digital twin of the welding process is formed that maintains data synchronization and interaction with the physical welding process.

4. The method for dynamic control of welding stress in metal components based on digital twins as described in claim 3, characterized in that, The instantaneous welding energy correction parameters calculated based on the welding energy dynamic control command include: When the energy reduction control command is received, the reduction amount of welding current or welding voltage is calculated by proportional integral algorithm based on the proportion by which the maximum tensile stress value exceeds the allowable tensile stress threshold, and used as an instant welding energy correction parameter. When the energy redistribution control command is received, the welding speed or welding torch oscillation amplitude is adjusted through a path planning algorithm based on the stress gradient distribution of the key area to generate instantaneous welding energy correction parameters for improving heat distribution. When the energy stabilization control command is received, the current welding parameters are set as the instant welding energy correction parameters.

5. The method for dynamic control of welding stress in metal components based on digital twins as described in claim 2, characterized in that, After adjusting the welding energy input, the updated welding signal is acquired again, and the digital twin of the welding process is driven to perform simulation verification. The verified stress field distribution includes: After performing welding energy adjustment based on the instantaneous welding energy correction parameters, and after a preset response delay time, the updated welding signal during the welding process is collected again by the multi-source sensor array. The updated welding signal is input into the welding process digital twin to update the heat input distribution, molten pool morphology, and microstructure in the welding process digital twin; In the updated digital twin of the welding process, a thermo-elastic-plastic finite element analysis was performed again to calculate the verified stress field distribution of the metal component after adjusting the welding energy input.

6. The method for dynamic control of welding stress in metal components based on digital twins as described in claim 5, characterized in that, The closed-loop calibration of the instantaneous welding energy correction parameters based on the verified stress field distribution includes: Calculate the residual between the verified stress field distribution and the expected stress field distribution; The validity of the instantaneous welding energy correction parameter is determined based on the magnitude and sign of the residual. If the residual exceeds the allowable range, the instantaneous welding energy correction parameter is calculated in reverse using a feedback control algorithm based on the residual to generate the calibrated welding energy parameter. If the residual is within the allowable range, then the instantaneous welding energy correction parameter is confirmed to be the calibrated welding energy parameter.

7. The method for dynamic control of welding stress in metal components based on digital twins as described in claim 1, characterized in that, The method combines the thermal strain field with the structural constraints and material constitutive relations in the digital twin of the welding process, and calculates the prototype stress field distribution inside the metal component under the action of welding thermal cycle by solving the mechanical equilibrium equations, including: The calculated thermal strain field is used as the initial strain load. The structural constraints formed by the fixture positioning points and rigid support points defined in the digital twin of the welding process are applied. The material constitutive relation is defined by the material yield strength, elastic modulus, and coefficient of thermal expansion as varying with temperature in the material property database. A set of mechanical equilibrium equations with nodal displacements as unknowns is established in the finite element solver, and geometric nonlinearity and material nonlinearity are considered for iterative solution. Finally, the prototype stress field distribution inside the metal component is output.

8. A dynamic stress control system for welding metal components based on digital twins, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the method for dynamic control of welding stress of metal components based on digital twins as described in any one of claims 1 to 7.

Images