Method for optimizing process parameters of multi-station high-speed laser welding of metal plates

Abstract

This invention provides a method for optimizing process parameters in high-speed laser welding of multi-station metal sheet metal parts, relating to industrial process control. It constructs a cross-station thermal state shift map to quantitatively characterize the transfer of residual heat from previous welding to subsequent welding areas, converting the residual thermal state into a laser heat input correction coefficient. Within the laser process parameter space, it performs collaborative mapping with heat input balance as a hard constraint and effective focal depth and spot coverage integrity as the boundary of the feasible defocusing domain. Combined with production line cycle time and equipment dynamic response capabilities, it executes a dynamic temporal arrangement, achieving feedforward compensation for heat accumulation effects between multiple stations and cross-station collaborative optimization of process parameters. This overcomes the limitations of traditional single-station calibration parameter translation and reuse that ignores inter-station thermal coupling effects, avoiding defects such as incomplete fusion or over-melting caused by subsequent stations using cold-state parameters due to initial temperature field shifts in the substrate. This improves the consistency of multi-station laser welding processes and the quality of weld formation.

Description

Technical Field

[0001] This invention relates to industrial process control, and more specifically, to a method for optimizing process parameters in high-speed laser welding of multi-station metal sheet metal parts. Background Technology

[0002] Laser welding technology, due to its high energy density, precise and controllable heat input, and excellent weld formation quality, has been widely used in high-end manufacturing fields such as automotive body manufacturing, aerospace structural components, and new energy battery modules. With the widespread application of lightweight body design and high-strength steel and aluminum alloys, multi-station laser welding production lines have become the mainstream process mode for welding large thin-walled metal sheet parts. In such production lines, the same workpiece needs to pass through multiple welding stations sequentially. Each station completes different weld tasks based on different combinations of sheet thickness, joint gaps, and heat dissipation conditions. Due to significant differences in the welding parameter combinations at each station, and the different fixture contact states, environmental heat dissipation conditions, and time intervals during the workpiece's transfer between stations, residual welding heat from previous stations will be transferred to the welding area of ​​subsequent stations through heat conduction mechanisms, causing a systematic shift in the initial temperature field of the substrate at subsequent stations relative to the single-station cold-state calibration conditions.

[0003] To address the optimization of laser welding process parameters, existing technologies mainly follow these paths: First, single-station experimental calibration methods. This involves designing orthogonal or response surface methodology experiments to establish a mapping relationship between process parameters and weld geometry, thereby determining the baseline parameter combinations for each station. While this method can obtain optimal parameters under specific conditions, it neglects the thermal coupling effect between multiple stations, treating each station as an independent welding process. Second, numerical simulation-based thermal field analysis methods. These methods use finite element or finite difference methods to simulate and calculate the welding temperature and stress fields. However, these methods are computationally intensive and time-consuming, making it difficult to meet the real-time requirements of production lines. Furthermore, they are often limited to single-pass welds or single-station scenarios, lacking a systematic modeling framework for the explicit characterization of heat accumulation and transfer between multiple stations. Third, online compensation methods based on temperature feedback. These methods collect pre-weld temperatures using infrared thermography or thermocouples and adjust the laser power of subsequent stations accordingly. While these methods can respond to changes in thermal state to some extent, they mostly employ empirical linear power reduction strategies, lacking a physical conversion mechanism from residual temperature rise to effective heat input deviation. Furthermore, sensor signals are susceptible to interference from smoke, splashes, and changes in surface condition, making it difficult to guarantee compensation accuracy and reliability. In addition, defocusing amount, a key optical parameter affecting melt depth and melt width, is often treated as a fixed value or an auxiliary adjustment in existing methods; no technical solution has yet been found to incorporate it as an active variable into a multi-station collaborative parameter optimization framework. Summary of the Invention

[0004] To address the aforementioned technical problems, this invention provides a method for optimizing process parameters in high-speed laser welding of multi-station metal sheet metal. This method addresses, to some extent, the problem in existing methods where single-station calibration parameters are transferred and reused across multiple stations, neglecting the impact of residual welding heat from previous stations on the initial temperature field of the substrate in subsequent stations, leading to deviations in the weld depth and width of subsequent stations from the target values.

[0005] According to one aspect of the present invention, a method for optimizing process parameters in high-speed laser welding of multi-station metal sheet parts is provided, comprising: The station flow information, spatial location information, and heat dissipation characteristics information of each station fixture of the same workpiece in the multi-station laser welding production line are obtained. The process of transferring residual heat from the welding of the previous station to the welding area of ​​the next station is segmented and corrected to generate a cross-station thermal state offset map. Based on the cross-station thermal state offset map and the weld quality target of each subsequent station, the residual thermal state of the preceding station is converted into a correction coefficient for the laser thermal input of the current station. Based on the correction coefficient, the parameter combinations of each station are collaboratively mapped in the laser process parameter space, and a cross-station collaborative parameter scheme is obtained with the goal of minimizing the weld formation quality deviation. Based on the cross-station collaborative parameter scheme, production line cycle time constraints, and equipment response capabilities, the switching sequence of parameters for each station is dynamically arranged, and an executable process parameter optimization sequence is output.

[0006] Furthermore, the step of generating the cross-station thermal state shift map includes: A thermal accumulation digital dictionary containing multiple working condition state vectors is pre-established. On the industrial side of the production line, the matching unit or interpolation unit of the dictionary is called according to the current working condition, and the residual heat temperature rise field across workstations is output using a reduced-order model. The working condition state vectors include at least plate characteristics, joint characteristics, fixture status, transfer time, and temperature status.

[0007] Furthermore, the reduced-order model is used to output the residual heat rise field across workstations, including determining the residual heat rise field of the area to be welded in the subsequent workstation as follows: in, Indicates the location of the area to be welded in the subsequent workstation j. The equivalent temperature rise at the location; This represents the nth reduced-order basis function extracted from a high-fidelity snapshot on the server side; Represents the working condition state vector The corresponding expansion coefficients; This is the overall correction factor for the online temperature rise field; The correction coefficient Update according to the following residual gating method: in, This represents the candidate correction value for the k-th batch; These are the residual gating coefficients; This represents the residual between the pre-welding temperature value and the model prediction value for the current batch; This represents the variance of the residuals within the sliding window; The stable scale is determined by the sensor noise level and the historical residual noise band.

[0008] Furthermore, the residual thermal state of the preceding station is converted into a correction coefficient for the laser heat input of the current station. The part that can effectively participate in the formation of the molten pool is screened from the residual temperature rise field through a spatial sensitivity function, and combined with temperature-related thermophysical properties, it is converted into residual heat energy. After slow feedback correction using the pre-welding temperature measurement residual as a gate, the correction coefficient that should be retained for the external laser heat input of the current station is calculated in reverse according to the proportion of the residual energy to the heat input required by the single station reference.

[0009] Furthermore, the residual thermal energy is calculated as follows: in, The corrected residual heat energy for welding segment q at station j. These are the model correction coefficients. For the effective volume to be welded, For temperature-dependent equivalent volumetric heat capacity, The spatial sensitivity function includes a transverse distribution with the weld centerline as the axis, a segmented distribution along the weld length, an attenuation distribution along the plate thickness direction, and an overlap interface correction factor.

[0010] Furthermore, the collaborative mapping takes the total heat input balance determined by the correction coefficient as a hard constraint. Within the feasible region jointly defined by the dual physical boundaries of decoking amount and the stable process window, with the goal of minimizing the comprehensive deviation of melt depth and melt width, the power, speed and decoking amount are collaboratively optimized and mapped through an evaluation function with a soft penalty term.

[0011] Furthermore, the coordinated solution for the power, speed, and defocus amount satisfies the following thermal input balance relationship: in, For laser power, For welding speed, This is the defocus amount. The length of the welded section. The effective absorption coefficient of the laser. The effective energy coupling factor corresponds to the defocusing amount and the characteristics of the current operating condition. To correct residual heat energy, The effective laser thermal input is used as a reference; and the defocusing amount is simultaneously constrained by the effective focal depth operating boundary and the operating boundary of the integrity of the spot coverage in the overlapping area.

[0012] According to another aspect of the present invention, a process parameter optimization system for high-speed laser welding of multi-station metal sheet metal parts is provided, comprising: The thermal state perception and map generation module is used to acquire the station flow information, spatial position information and heat dissipation characteristic information of each station fixture of the same workpiece in the multi-station laser welding production line. It performs segmented correction on the transfer process of residual heat from the welding of the previous station to the welding area of ​​the subsequent station and generates a cross-station thermal state offset map. The heat input correction coefficient calculation module is used to convert the residual heat state of the previous station into the correction coefficient of the laser heat input of the current station based on the cross-station heat state offset map and the weld quality target of each subsequent station. The parameter collaborative mapping module is used to collaboratively map the parameter combinations of each station in the laser process parameter space according to the correction coefficient, and obtain a cross-station collaborative parameter scheme with the goal of minimizing the weld formation quality deviation. The timing arrangement and control output module is used to dynamically arrange the switching timing of parameters of each workstation according to the cross-workstation collaborative parameter scheme, production line cycle time constraints and equipment response capabilities, and output an executable process parameter optimization sequence.

[0013] Compared with existing technologies, the process parameter optimization method for high-speed laser welding of multi-station metal sheet metal provided by this invention realizes the quantitative transfer characterization of residual heat from the previous welding to the subsequent welding area by constructing a cross-station thermal state offset map, and converts the residual thermal state into a laser thermal input correction coefficient accordingly. Then, it performs collaborative mapping in the three-dimensional process parameter space composed of laser power, welding speed and defocusing amount with thermal input balance as a hard constraint. Finally, it performs time-series dynamic arrangement in combination with production line cycle time and equipment dynamic response capability, thereby realizing feedforward compensation of heat accumulation effect between multiple stations and cross-station collaborative optimization of process parameters. This effectively overcomes the technical limitations of traditional single-station test calibration methods that ignore the thermal coupling effect between stations, and avoids defects such as incomplete fusion or over-fusion caused by the direct use of single-station cold-state parameters in subsequent stations due to shifts in the initial temperature field of the substrate. This significantly improves the consistency of multi-station laser welding processes and the quality of weld formation. Simultaneously, by introducing defocusing as an active adjustment variable and constraining its feasible region with dual physical boundaries of effective focus depth and spot coverage integrity, coupled with a soft penalty evaluation mechanism based on a dynamic process database, the optimized process parameters always remain within the calibrated and verified feasible process region. This avoids the risk of response surface model extrapolation and empowers the system with... The invention offers the ability to progressively explore unknown operating conditions and adaptively expand boundaries. Furthermore, by using pre-welding online temperature measurement data as the criterion for residual gating instead of directly correcting the input, it effectively suppresses the interference of welding spatter, fumes, and surface condition changes on the sensing signals, ensuring the robustness of thermal state perception and parameter correction. In addition, by deploying high-fidelity thermal accumulation solutions on the server side and degraded model calls on the production line's industrial PC side in a layered architecture, along with online learning of the equipment's recent response characteristics and dynamic updates of transition time margins, it unifies offline physical accuracy with online computational real-time performance, enabling complex algorithms to run stably within production line cycle time constraints. Therefore, this invention comprehensively improves the process adaptability, process stability, and product quality consistency of multi-station laser welding production lines under varying operating conditions, providing reliable process optimization and control technologies for high-quality welding manufacturing of large metal sheet parts such as automotive bodies. Attached Figure Description

[0014] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. In the drawings: Figure 1 This is a flowchart illustrating a method for optimizing process parameters in high-speed laser welding of multi-station metal sheet metal according to an embodiment of the present invention.

[0015] Figure 2This diagram illustrates an application scenario of the process parameter optimization method for high-speed laser welding of multi-station metal sheet metal according to an embodiment of the present invention. Detailed Implementation

[0016] Hereinafter, exemplary embodiments according to the present invention will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are merely some embodiments of the present invention, and not all embodiments of the present invention. It should be understood that the present invention is not limited to the exemplary embodiments described herein.

[0017] As mentioned in the background section above, current production processes typically rely on single-station test calibration and then reuse the parameters, neglecting the changes in local heat input state caused by heat accumulation and transfer between multiple stations. The residual welding heat from the preceding station will cause the initial temperature field of the substrate in the subsequent station to shift. If the original parameter combination is still used, the penetration depth and penetration width of the subsequent station will deviate from the target value, resulting in defects such as incomplete fusion or over-fusion. Existing methods lack explicit characterization of the heat accumulation and transfer effect between stations and a mechanism for cross-station parameter collaborative correction.

[0018] Figure 1 This is a flowchart illustrating a method for optimizing process parameters in high-speed laser welding of multi-station metal sheet metal according to an embodiment of the present invention. Figure 1 As shown, the process parameter optimization method for high-speed laser welding of multi-station metal sheet metal includes: S1. The process planning calculation unit generates a high-fidelity thermal accumulation digital dictionary before the server during the shift or batch offline planning stage. Then, on the industrial PC side of the production line, dictionary matching, local interpolation, and low-dimensional expansion are performed based on the current working conditions to output a cross-station thermal state offset map.

[0019] First, the process planning calculation unit reads multi-station welding sequence information from the production line process database. This multi-station welding sequence information includes at least the workpiece number, station number, weld number, welding start time, welding completion time, workpiece departure time, entry time to the next station, transfer path length between adjacent stations, workpiece posture change information, and the spatial position of each station's weld in the workpiece coordinate system. Then, it reads the laser power, welding speed, defocusing amount, weld length, plate thickness combination, joint gap, plate grade, and effective laser absorption coefficient corresponding to each station from the welding process parameter database. Finally, it reads the contact area, fixture material, cooling state, clamping position, equivalent contact heat transfer coefficient, convection heat transfer calibration coefficient, surface emissivity, and pre-welding ambient temperature record of each station's fixture from the fixture structure database, fixture thermal calibration database, and equipment calibration database. The equivalent contact heat transfer coefficient is used to characterize the heat conduction and heat dissipation capacity when the workpiece is in contact with the fixture, the convective heat transfer calibration coefficient is used to characterize the convective heat dissipation capacity when the workpiece is exposed to air or a protective gas environment, and the surface emissivity is used to characterize the radiative heat dissipation capacity under high temperature conditions.

[0020] Furthermore, the server-side employs a high-fidelity method to pre-solve thermal accumulation snapshots for different plate thickness combinations, joint types, fixture states, and temperature ranges. This high-fidelity method includes temperature-related thermal properties, double-ellipsoidal or Gaussian heat sources, multimodal heat dissipation boundaries, and actual workpiece posture changes. Based on this, reduced-order basis functions and local response parameters are extracted to form a thermal accumulation digital dictionary. This digital dictionary uses a working condition state vector as an index, which at least includes plate thickness combinations, joint gaps, fixture states, transfer time, pre-welding temperature range, surface emissivity, and equivalent comprehensive heat dissipation parameters. The server-side high-fidelity solution can utilize finite element method, finite volume method, or high-precision integration to generate thermal snapshot fields for representative working conditions. Several spatial basis functions are then formed through principal component decomposition, truncated singular value decomposition, or other reduced-order basis extraction methods. The corresponding expanded structure is used. On the industrial PC side, only dictionary matching, interpolation, and low-dimensional expansion are performed; the complete transient heat transfer equation is not directly resolved.

[0021] Furthermore, to avoid ambiguity in determining whether the dictionary covers the current operating condition, the system introduces a coverage determination mechanism during online calls. Specifically, this mechanism uses the current operating condition state vector... For the query point, calculate its weighted Euclidean distance to all snapshot state vectors in the dictionary. The weights are determined by the sensitivity of plate thickness, joint gap, fixture status, transfer time, and pre-welding temperature to changes in thermal state. If the minimum distance is less than the coverage radius... If the minimum distance is within the coverage area, it is determined to fall within the coverage area, and interpolation is used for output; if the minimum distance is between With edge radius If the minimum distance is between [a certain value], it is considered an edge case, and the nearest neighbor superposition conservative margin output is used; if the minimum distance is greater than [a certain value], it is considered an edge case. If the condition is not met, it is determined to be outside the coverage area, a limited version is output, and an offline supplementary snapshot queue is triggered. and The prediction error distribution is determined by the validation sample set, and is not based solely on subjective human values.

[0022] Furthermore, to reduce errors caused by constantting thermal property parameters, the system reads calibration curves of plate density, specific heat capacity, and thermal diffusivity as a function of temperature from the material thermal property database. For the heat transfer correlation between each preceding and succeeding station, the system calculates the equivalent volumetric heat capacity and equivalent thermal diffusivity within that temperature range based on the estimated post-weld temperature range and the pre-weld temperature baseline. If the material thermal property database records segmented temperature curves, the system reads the corresponding parameters according to the temperature range where the current residual temperature rise is located; if only discrete material sample points are available, the system interpolates between adjacent sample points. The temperature-related thermal property parameters are obtained from material handbooks, supplier test reports, or production line thermal response calibration, and are stored in association with the plate grade and thickness combination.

[0023] Furthermore, for the near-weld region, the system does not simply equate the preceding heat source to a point heat source. Instead, it reads the optical calibration and welding heat source calibration results and uses surface or volume heat sources distributed along the weld centerline to represent the preceding heat input. This heat source distribution can be represented by a Gaussian surface heat source, a double ellipsoidal heat source, or an off-field heat source core calibrated by infrared thermography. For the area to be welded that is far from the preceding weld and has a gentle spatial temperature gradient, a line source approximation can be used to reduce computational load. Thus, the system retains the geometric distribution of heat sources in the region near the preceding weld and the lap interface, while using a simplified heat transfer core in the far-field region, maintaining a balance between computational accuracy and industrial computing efficiency.

[0024] In one implementation, the system meshes the welding area of ​​subsequent workstations, with each mesh cell possessing unique spatial coordinates, cell volume, and the identifier of the welding segment to which it belongs. For each subsequent mesh cell, the system traverses its set of workstations contributing residual heat and superimposes the expanded result of matching the residual heat temperature rise field of the preceding workstation with the current working condition state vector to obtain the residual thermal state of that mesh cell. The cross-workstation residual thermal state offset can be determined according to the following formula: in, Indicates the location of the area to be welded in the subsequent workstation j. The equivalent temperature rise caused by residual heat from the preceding workstation, expressed in K; This represents the nth reduced-order basis function extracted from a high-fidelity snapshot on the server side; Represents the working condition state vector The corresponding expansion coefficient, the working condition state vector includes at least plate thickness combination, joint gap, fixture state, transfer time, pre-welding temperature range, surface emissivity and equivalent comprehensive heat dissipation parameters; It is the overall correction coefficient for the online temperature rise field, a dimensionless quantity, used to slowly absorb continuous and stable batch deviations; This represents the candidate correction value for the k-th batch; For residual gating coefficients, This represents the residual between the pre-welding temperature value and the model prediction value for the current batch. This represents the variance of the residuals within the sliding window. The stability scale is determined jointly by the sensor's factory calibration noise and historical residual noise band. Because... As the residual variance changes continuously, when the temperature measurement residual is stable, Approaching 1, Slowly follow updates; when the temperature residual changes drastically or the sensor signal is interfered with. Approaching 0 Approximate freezing is used to avoid directly transmitting noise into the thermal state migration map.

[0025] Furthermore, the system performs the initialization phase... Perform a cold start setup. Specifically, before the system accumulates at least 20 batches of stable residual samples, Take three times the sensor's factory-calibrated noise level as a conservative benchmark; after accumulating at least 20 batches of stable residual samples, The system switches to an unbiased estimate of the historical residual standard deviation and updates it batch by batch using a sliding window approach. This process allows the system to adopt a conservative decision boundary during the cold start phase and gradually transition to data-driven adaptive updates after sample accumulation, avoiding overly fast or slow corrections due to insufficient initial samples.

[0026] Furthermore, for the welding area of ​​each subsequent workstation, the system selects interpolated output, nearest neighbor conservative output, or restricted version output based on the coverage determination result. If the current working condition falls within the dictionary coverage range, the interpolated expansion coefficient is used for calculation; if it is an edge working condition, a conservative margin is added to the nearest neighbor expansion result to avoid extrapolation errors; if it exceeds the coverage range, the restricted version of the thermal state offset result is output, and the working condition is written to the offline supplementary snapshot queue. The resulting cross-workstation thermal state offset map can simultaneously meet the requirements of online real-time performance and engineering conservatism.

[0027] After all the mesh elements to be welded at all subsequent workstations have been calculated, the system writes the coordinates of the area to be welded at each subsequent workstation, the thermal contribution components of each preceding workstation, the superimposed residual heat equivalent temperature rise, heat dissipation segment information, temperature range of thermophysical parameters, heat source distribution type, online feedback correction coefficient, and spatial overlap relationship into the cross-workstation thermal state offset map. The cross-workstation thermal state offset map can be stored in the form of a two-dimensional table, three-dimensional mesh data, or key-value pairs. Each entry is indexed by the workstation number, welding segment number, and spatial coordinates, and its attribute values ​​are the residual heat equivalent temperature rise and contribution source. After the map is generated, the system stores it in the data cache of the process parameter optimization system as input for the calculation of the substrate initial temperature field inversion correction and heat input correction coefficient in S2.

[0028] S2, the heat input correction calculation unit reads the cross-station thermal state offset map generated by S1, and combines it with the weld structure, target penetration depth, target weld width, single-station reference effective heat input, and pre-welding online temperature measurement data of each subsequent station to perform inversion correction on the initial temperature field of the substrate in the area to be welded in the subsequent station, and obtains the heat input correction coefficient of the welding section of each station.

[0029] First, the heat input correction calculation unit reads the coordinates of the target station's welding area, the residual heat equivalent temperature rise, and the heat contribution components of each preceding station from the cross-station thermal state offset map. Then, it reads the weld length, plate thickness combination, lap width, joint gap distribution, target penetration depth, target weld width, reference laser power, reference welding speed, reference defocusing amount, and reference laser effective absorption coefficient from the weld process database. Finally, based on the target weld width, plate thickness combination, and weld length, it determines the effective welding volume for each welding segment of the station. This effective welding volume is not the entire workpiece volume, but rather the local volume within the current welding segment that can participate in the formation of the molten pool and influence the penetration depth and weld width, including the area covered by the laser spot, the area near the lap interface, and the area near the target penetration path.

[0030] Furthermore, the system establishes a spatial sensitivity function for each welding segment at each workstation. This spatial sensitivity function describes the degree to which the residual temperature rise at different locations in the welding area contributes to the formation of the current molten pool. Instead of using a single empirical weight, it employs a calibrable family of functions. Specifically, this family of functions uses the weld centerline as an axis, employs a Gaussian distribution laterally, a piecewise uniform distribution along the length, and a decaying distribution along the plate thickness, with an additional lap joint correction factor. Its parameters are obtained through inversion calibration using metallographic samples of the same plate thickness combination and joint type from the post-weld cross-section of a single workstation. If the incoming material batch, surface condition, or fixture heat dissipation status deviates from the original calibration sample under the current working condition, the system interpolates or slightly updates the function family parameters based on historical sampling results and pre-weld temperature measurements. If the current working condition exceeds the coverage of the existing function family, the most recently covered parameters are used and written into the offline supplementary calibration queue, rather than being directly extrapolated.

[0031] Furthermore, the pre-welding online temperature measurement data is not directly used as the heat input reduction factor in the physical formula, but rather as an abnormal state trigger and a slow correction source. The system reads the temperature data collected by the infrared temperature measurement unit or thermocouple and converts it into the workpiece surface temperature or clamping point temperature according to the sensor calibration curve; then, the coordinates of the temperature measurement point are mapped to a unified workpiece coordinate system, and the sliding statistics of the temperature measurement residual and the model prediction residual are calculated. The model correction coefficients are only allowed to be updated slowly when the residual variance is continuously within the historical noise band and the residual direction is stable; once the residual jumps sharply, the surface emissivity fluctuates significantly, or the temperature measurement signal is interfered with by splashes, smoke, or oil, the system immediately freezes the correction coefficients, fully trusts the physical prediction results, and does not allow false high temperature signals to directly reduce subsequent power.

[0032] In one embodiment, the residual thermal energy and the heat input correction coefficient can be determined according to the following formula: in, This represents the residual thermal energy after correction by the space sensitivity function and online feedback, in J; These are model correction coefficients, which are dimensionless quantities used to absorb continuous and stable operating condition deviations. This represents the effective volume to be welded in section q, in meters (m). 3 ; This indicates the current initial temperature of the board material. Equivalent volumetric heat capacity over the corrected temperature range, in J / (m³) 3 ·K), can be obtained by integral averaging of temperature-dependent density and specific heat capacity curves from a materials database; Indicates that calibration parameters are included. The space sensitivity function is a dimensionless quantity; This represents the residual heat equivalent temperature rise of S1 output, in K. Let be the model correction coefficients for the k-th batch. The candidate correction values ​​are obtained by back-substituting the residuals of the current batch. The residual gating coefficients, based on the same concept as S1, are determined jointly by the variance of the temperature measurement residuals within the sliding window and the sensor noise scale. Because... The correction coefficient is updated slowly when the temperature measurement residual is stable and frozen when the temperature measurement residual is abnormal, thus avoiding sensor noise from contaminating the thermal input correction results.

[0033] Furthermore, the system corrects for residual thermal energy based on temperature-dependent thermal properties, rather than treating density, specific heat capacity, and thermal diffusivity as constants. For different plate thickness combinations and different temperature ranges, the system reads piecewise thermal property curves from the material database, calculates the equivalent volumetric heat capacity within the current correction temperature range, and writes it into the heat input correction data table. For welded sections located in areas with plate thickness spans, local changes in joint gaps, or switching of fixture heat dissipation states, the system uses piecewise integration instead of single-point conversion to reduce errors caused by local thermal cycling. Through this processing, residual thermal energy is no longer just a simple product of temperature rises, but rather an effective heat input deviation constrained by both temperature-dependent thermal properties and spatial sensitivity functions.

[0034] Furthermore, to ensure The cold start is executable; the system will execute during the initialization phase. Set to three times the sensor's factory-calibrated noise; after accumulating at least 20 batches of stable residual samples, The algorithm switches to an unbiased estimate of the historical residual standard deviation and updates it batch by batch using a sliding window approach. Through this process, S2 initially employs conservative gating and then gradually transitions to data-driven adaptive updates, avoiding overly rapid corrections when samples are insufficient.

[0035] Once all welding segments at the current workstation have completed the calculation of residual heat energy and heat input correction coefficients, the system creates a heat input correction data table. This data table, indexed by workstation number and welding segment number, records the corrected initial temperature field of the substrate, corrected residual heat energy, reference effective laser heat input, heat input correction coefficient, spatial sensitivity function parameters, temperature data source, online feedback correction coefficient, and whether the current working condition falls within the calibration envelope range in the corresponding entries, and then transmits this information to the parameter collaborative mapping unit of S3.

[0036] S3, the parameter collaborative mapping unit reads the heat input correction coefficient and corrected residual heat energy output by S2, and within the parameter space composed of laser power, welding speed and defocusing amount, it reconstructs the cross-station collaborative parameter scheme of each station under the constraints of heat input balance after residual heat compensation, dual physical boundaries of defocusing amount, dynamic calibration envelope and stable welding process window.

[0037] First, the parameter collaborative mapping unit reads the corrected residual heat energy, reference effective heat input, and heat input correction coefficient for the target station and target welding segment from the heat input correction data table. Then, it reads the laser's stable output power range, the robot or machine tool's achievable welding speed range, the defocus adjustment mechanism's achievable position range, power response capability, speed response capability, and defocus axis positioning capability from the equipment capability database. Next, it reads the correspondence between defocusing amount and spot radius, energy density distribution, keyhole stability, plasma shielding characteristics, and effective energy coupling factor from the optical calibration database and dynamic process database. Simultaneously, the system reads the plate thickness combination, overlap width, joint gap distribution, target weld depth, target weld width, and defect control requirements for the target welding segment.

[0038] Furthermore, the defocusing amount is treated as an actively adjusted variable in this step. The system determines the effective focal depth boundary based on the optical system's focal depth, the surface height fluctuation of the plate, and the plate thickness combination; simultaneously, it determines the spot coverage boundary based on the calibration curves of the overlap width, joint gap, and spot radius as a function of the defocusing amount, as well as the weld center positioning deviation. If the spot energy distribution corresponding to a certain defocusing amount can cover the overlap interface and still meet the effective focal depth conditions required for the formation of the target melt depth, then this defocusing amount is added to the defocusing amount candidate set; if a certain defocusing amount causes the main energy distribution to deviate from the overlap area or fails to cover the effective melt depth formation area in the plate thickness direction, then this defocusing amount is not included in the subsequent solution. Thus, the defocusing amount candidate set is formed by both optical and geometric constraints, rather than by a fixed range of values ​​specified manually.

[0039] Furthermore, the system reads a set of calibration samples from the dynamic process database corresponding to the current plate thickness combination, joint type, fixture heat dissipation status, and joint gap range. This dynamic process database is continuously updated during production line debugging and mass production sampling. Each sample set includes at least the following parameters: laser power, welding speed, defocusing amount, effective heat input per unit length, effective energy coupling factor, plate thickness combination, joint gap, pre-weld temperature, fixture heat dissipation status, post-weld penetration depth, post-weld weld width, porosity status, spatter status, surface forming status, and cross-sectional quality results. The system determines the calibration envelope based on the position of the current target condition in the sample space. The current candidate parameter combination only participates in the solution within the calibration envelope or the interpolation region of adjacent samples, avoiding extrapolation of the response relationship within unverified regions. For parameter combinations falling into the unstable transition zone between deep penetration welding and thermal conductivity welding, the porosity concentration zone, the spatter abnormal zone, or the cross-sectional quality unstable zone, the system does not directly classify them as failures but instead transfers them to a soft penalty region for further evaluation.

[0040] In one implementation, the coordinated mapping of power, speed, and defocus amount can be determined based on the following core constraints: in, Let q represent the laser power to be solved for welding segment q at station j, in W. This represents the welding speed to be solved, in m / s; This represents the defocusing amount to be determined, in mm; This indicates the length of the welded section, in meters (m). This represents the effective laser absorption coefficient at the current workstation, which is a dimensionless parameter. The effective energy coupling factor, representing the combined effective energy coupling factor of the candidate defocus amount and the current operating condition characteristics, is a dimensionless quantity. The operating condition feature vector... At least including plate thickness combination, joint gap, pre-welding temperature, spot radius, keyhole stability characteristics, and plume monitoring characteristics; therefore The unit is J, which represents the external heating input provided by the laser and effectively involved in the formation of the molten pool under the current parameter combination. This represents the corrected residual heat energy obtained from S2, expressed in J. This represents the effective laser thermal input required to achieve the target melt depth and width under single-station benchmark calibration conditions, expressed in J. Both sides of the equation are in the dimension of energy, indicating that the total effective thermal input required for the formation of the molten pool in subsequent stations consists of the external laser thermal input of the current station and the residual thermal input of the preceding station. For the evaluation function with soft penalties, and These are the melt depth and melt width predicted by the process response relationship, in mm; and These are the target weld penetration and target weld width for the weld segment, respectively, in mm; and The weights are dimensionless. This represents the smoothed distance penalty from the current calibration envelope. This represents a smooth distance penalty from the stable welding process window. The penalty increases continuously as the candidate parameter deviates from the calibration envelope or stable window, rather than jumping directly to infinity at the boundary, thus allowing the system to be engineered using a gradual, iterative approach rather than abrupt shutdown.

[0041] The above formula allows both heat input balance and soft constraint evaluation to participate in the parameter solution simultaneously. When the residual heat energy of the preceding station increases, the system will not simply apply the single-station reference power and reference speed, but will instead ensure... The external laser heat input is recalculated based on the total effective heat input. Simultaneously, candidate parameters are preferentially searched within the calibration envelope and stable welding window. If all feasible combinations are close to the boundary, the system does not directly declare no solution, but instead generates short-segment trial parameters that only apply to the initial short zone of the weld. The system collects molten pool radiation, reflection, hole depth, or weld penetration proxy signals through an online monitoring unit. If the monitoring results indicate that the keyhole is stable, without significant porosity or spatter, and the weld penetration evolution is smooth, the combination is upgraded from the trial state to the usable state, and the corresponding boundary is conditionally expanded outwards. If the monitoring results show that the keyhole closure is unstable, the defect trend is enhanced, or the cross-sectional deviation is expanded, the system immediately reverts to a more conservative parameter combination instead of stopping the adaptive process.

[0042] Furthermore, for each candidate defocus amount, the system calculates the condition based on the thermal input balance relationship. Required The ratio is then used to traverse the equipment's executable power range and welding speed range, filtering out candidate power-speed combinations that satisfy the ratio relationship and fall within the calibration envelope and stable welding process window. For each candidate power-speed combination, the system calls the welding process response relationship to predict its corresponding penetration depth and weld width. This welding process response relationship can be established offline during the production line commissioning phase for a limited number of plate thickness combinations, joint gap ranges, and fixture heat dissipation conditions involved in this solution, and dynamically updated during mass production based on sampling inspection results.

[0043] Specifically, the system employs orthogonal or uniform experimental design to acquire multiple sets of welding samples within the executable range of laser power, welding speed, and defocusing amount. Each set of samples records the effective heat input per unit length, defocusing amount, effective energy coupling factor, plate thickness combination, joint gap, pre-weld temperature, and the weld penetration and weld width obtained from post-weld cross-sectional inspection. Subsequently, the system uses the effective heat input per unit length and defocusing amount as the main independent variables, and plate thickness combination, joint gap, and pre-weld temperature as grouping identifiers or correction terms. It then uses the least squares method to fit a binary quadratic response surface, or, when the sample density is high, uses piecewise spline interpolation to form the welding process response relationship. After fitting, the system uses validation samples not involved in the fitting to calculate the residual distribution of the predicted weld penetration and predicted weld width. When the residual distribution meets the allowable band requirements for product weld quality, the response relationship is written into the calibration library for the corresponding plate thickness combination and joint type for subsequent parameter collaborative mapping.

[0044] After the predicted weld depth and weld width of the candidate combinations are calculated, the system compares the predicted weld depth with the target weld depth and the predicted weld width with the target weld width. During the comparison, the weld depth deviation is normalized to the target weld depth or the product quality allowable zone, and the weld width deviation is normalized to the target weld width or the effective overlap width, thus avoiding the direct addition of data with different dimensions and scales. If a candidate combination falls within the product quality allowable zone in both weld depth and weld width dimensions, and does not belong to the area of ​​porosity, spatter, or unstable cross-section samples, the system adds it to the feasible parameter set. If multiple candidate combinations are added to the feasible parameter set, the system further compares the variation range between the combination and the parameters of adjacent stations, prioritizing combinations where power variation, speed variation, and defocusing variation are more conducive to smooth switching. If no candidate combination simultaneously satisfies the requirements for weld depth, weld width, equipment operating boundary, calibration envelope, and stable welding window, the system writes the infeasibility status of the current station and its limiting reasons into the parameter solution record to adjust the baseline process window, supplement calibration samples, or correct fixture heat dissipation conditions.

[0045] Furthermore, for candidate parameter combinations near the boundary of the stable welding process window, the system does not directly determine them as infeasible, but instead generates short-segment trial parameters. The length of the trial segment does not exceed 8% or 30mm of the total weld length, and is only activated within this length range. The online monitoring unit collects molten pool radiation, reflection, or hole depth proxy signals at a sampling rate of not less than 2kHz. If no stable keyhole signal is detected within 100ms after the start of the trial segment, or if the weld depth growth rate is lower than the dynamic judgment boundary, the trial segment is immediately terminated, subsequent welding is paused, and the system reverts to the conservative parameter combination verified before this trial. At the same time, this trial condition is recorded as a state to be supplemented for calibration. If the trial result indicates that the condition is stable, the boundary corresponding to the candidate parameter combination is conditionally expanded outward and written into the dynamic process database. Through this process, the system can avoid unsolvable shutdowns caused by an overly tight calibration envelope, and can gradually expand the process coverage range while meeting safety requirements.

[0046] After all welding segments within the target station have completed parameter solving, the system creates a cross-station collaborative parameter scheme data structure. This data structure uses the station number and welding segment number as indexes, and writes the target laser power, target welding speed, target defocusing amount, corrected residual heat energy, reference effective laser heat input, external effective laser heat input, predicted weld depth, predicted weld width, effective focus depth state, spot coverage state, calibration envelope state, stable welding window state, and parameter changes of adjacent stations into corresponding entries. After all stations have been processed, the system sends the cross-station collaborative parameter scheme to the timing arrangement and control output unit of S4 to generate a parameter optimization sequence that can be executed by the production line control system.

[0047] S4, the timing arrangement and control output unit reads the cross-station collaborative parameter scheme output by S3, and combines it with the transfer time between adjacent stations, the production line cycle time distribution relationship, the dynamic response capability of the equipment, the recent execution feedback and the synchronization error record of multiple controllers to dynamically arrange the switching timing of laser power, welding speed and defocusing amount of each station, and output the multi-station laser welding process parameter optimization sequence.

[0048] First, the system reads the workpiece transfer time, fixture opening time, handling time, repositioning time, fixture closing time, and pre-welding stabilization time between adjacent workstations from the production line cycle time database. Then, it reads the laser power ramp-up capability, power drop-down capability, robot or machine tool speed planning capability, motion axis acceleration capability, motion axis jerk capability, defocus adjustment mechanism displacement speed, defocus adjustment mechanism acceleration, and defocus axis positioning stabilization time from the equipment response database. Next, it reads the actual laser power response curve, robot or machine tool actual speed curve, defocus axis actual position curve, controller communication delay record, clock synchronization deviation record, and response change record under different load conditions from the recent execution feedback database. Finally, it reads the target laser power, target welding speed, and target defocus amount for each workstation or welding segment from the cross-workstation collaborative parameter scheme output by S3.

[0049] Furthermore, the system calculates parameter changes for each pair of adjacent workstations. For laser power, the system calculates the difference between the target power of the preceding and following workstations, and determines the time required for a smooth power transition based on the laser's recent actual response curve and calibration response curve. For welding speed, the system calculates the difference between the target speed of the preceding and following workstations, and determines the time required for a speed transition based on the robot's or machine tool's speed planning capability under corresponding path posture, load conditions, and shaft temperature conditions. For defocusing, the system calculates the positional difference between the target defocusing amounts of the preceding and following workstations, and determines the time required for defocusing switching and stabilization based on the recent positional response data of the defocusing adjustment mechanism. These times do not depend solely on static factory calibration, but are determined jointly by equipment calibration data and recent actual execution feedback.

[0050] Furthermore, the system determines the available parameter transition time based on the transfer time between adjacent workstations and the production line cycle time allocation ratio. The production line cycle time allocation ratio can be obtained from the production line process formula, cycle time balancing results, or continuous stable production records, and is used to represent the proportion of time available for parameter preparation and smooth parameter switching during the workpiece transfer process between adjacent workstations. The system also forms a transition time margin based on recent communication delays, clock synchronization deviations, and equipment response fluctuation distributions. This transition time margin originates from the distribution characteristics of actual execution data, rather than being a fixed setting. The system matches the power transition time, speed transition time, defocusing transition time, and transition time margin together with the available parameter transition time. When the transition process of the three types of parameters can be completed within the available parameter transition time, the system arranges the switching processes of power, speed, and defocusing amount within the workpiece transfer, pre-welding positioning, or pre-welding stabilization stages. When the transition process of any parameter cannot be completed within the available parameter transition time, the system does not directly stop the machine but instead feeds back to S3 to reselect an alternative combination with a smaller change amplitude that still meets the soft penalty evaluation.

[0051] During the specific control curve generation process, the system generates a smooth power curve for laser power, ensuring that the laser reaches the target power or target standby state before entering the next welding station; it generates a speed transition curve with limited jerk for welding speed, ensuring that the robot or machine tool reaches the target welding speed before reaching the welding start point of the next station; and it generates a continuous displacement curve for defocusing, retaining the defocusing axis stabilization time before the laser is turned on. For control systems supporting multi-channel synchronous interpolation, the system maps the power curve, speed curve, and defocusing curve to the same time axis and compensates for the command timestamp based on the controller clock synchronization deviation. For production lines where the laser, robot, and defocusing adjustment mechanism are executed by different controllers, the system uses the workpiece arrival signal, fixture closure signal, and laser enable signal as synchronization references to time-align the commands of each controller, and advances or delays the transmission time of the corresponding commands based on the historical communication delay distribution.

[0052] On the other hand, if recent feedback indicates a persistent drift in equipment response, the system only updates the transition time margin and curve slope for subsequent batches or shifts, without forcibly reverting issued instructions in currently executed workstations. If abnormal jumps occur in temperature measurement, power feedback, or displacement feedback, the anomaly is marked as sensor interference or communication disturbance, the current online calibration is frozen, and only the physical baseline plan is retained. Fine-tuning resumes only after the anomaly is resolved. In this way, S4 uses the high-fidelity planning results as a baseline and only implements minor corrections under the premise of stable actual response, thereby avoiding over-parameter tuning due to real-time noise.

[0053] Once the parameter switching sequences for all adjacent workstations are arranged, the system creates a multi-workstation laser welding process parameter optimization sequence. This optimization sequence uses the workpiece number, workstation number, and welding segment number as indexes, and writes the welding start time, welding end time, target laser power, target welding speed, target defocusing amount, parameter transition start time, parameter transition end time, power smoothing curve, speed smoothing curve, defocusing smoothing curve, inter-workstation transfer time, cycle time allocation ratio, equipment response parameter source, synchronization compensation information, transition time margin source, and corresponding heat input correction coefficient into a structured data table. Subsequently, the system distributes this optimization sequence to the production line control system, robot controller, laser controller, and defocusing adjustment mechanism, ensuring that when the same workpiece moves between multiple welding workstations, the welding parameters for each workstation are jointly determined based on the cross-workstation residual heat state, subsequent heat input correction results, dynamic calibration database constraints, and the equipment's executable switching sequence.

[0054] In summary, the process parameter optimization method for high-speed laser welding of multi-station metal sheet metal based on the embodiments of the present invention has been clarified. It realizes the quantitative transfer characterization of residual heat from the previous welding to the subsequent welding area by constructing a cross-station thermal state offset map, and converts the residual thermal state into a laser heat input correction coefficient accordingly. Then, it performs collaborative mapping in the three-dimensional process parameter space composed of laser power, welding speed and defocusing amount with heat input balance as a hard constraint. Finally, it performs time-series dynamic arrangement by combining the production line cycle time and equipment dynamic response capability, thereby realizing feedforward compensation of heat accumulation effect between multiple stations and cross-station collaborative optimization of process parameters. This effectively overcomes the technical limitations of traditional single-station test calibration methods that ignore the thermal coupling effect between stations, and avoids defects such as incomplete fusion or over-fusion caused by the direct use of single-station cold-state parameters in subsequent stations due to shifts in the initial temperature field of the substrate. This significantly improves the consistency of multi-station laser welding processes and the quality of weld formation. Simultaneously, by introducing defocusing as an active adjustment variable and constraining its feasible region with dual physical boundaries of effective focus depth and spot coverage integrity, coupled with a soft penalty evaluation mechanism based on a dynamic process database, the optimized process parameters always remain within the calibrated and verified feasible process region. This avoids the risk of response surface model extrapolation and empowers the system with... The invention offers the ability to progressively explore unknown operating conditions and adaptively expand boundaries. Furthermore, by using pre-welding online temperature measurement data as the criterion for residual gating instead of directly correcting the input, it effectively suppresses the interference of welding spatter, fumes, and surface condition changes on the sensing signals, ensuring the robustness of thermal state perception and parameter correction. In addition, by deploying high-fidelity thermal accumulation solutions on the server side and degraded model calls on the production line's industrial PC side in a layered architecture, along with online learning of the equipment's recent response characteristics and dynamic updates of transition time margins, it unifies offline physical accuracy with online computational real-time performance, enabling complex algorithms to run stably within production line cycle time constraints. Therefore, this invention comprehensively improves the process adaptability, process stability, and product quality consistency of multi-station laser welding production lines under varying operating conditions, providing reliable process optimization and control technologies for high-quality welding manufacturing of large metal sheet parts such as automotive bodies.

[0055] Here, those skilled in the art will understand that the specific operations of each step in the above-mentioned multi-station high-speed laser welding process parameter optimization system for metal sheet parts have been referenced above. Figure 1 and Figure 2 The process parameter optimization method for high-speed laser welding of multi-station metal sheet metal has been described in detail, and therefore, its repeated description will be omitted.

[0056] In summary, the process parameter optimization system for high-speed laser welding of multi-station metal sheet metal based on the embodiments of the present invention has been clarified. It realizes the quantitative transfer characterization of residual heat from the previous welding to the subsequent welding area by constructing a cross-station thermal state offset map, and converts the residual thermal state into a laser heat input correction coefficient accordingly. Then, it performs collaborative mapping in the three-dimensional process parameter space composed of laser power, welding speed and defocusing amount with heat input balance as a hard constraint. Finally, it performs time-series dynamic arrangement by combining the production line cycle time and the dynamic response capability of the equipment, thereby realizing feedforward compensation of heat accumulation effect between multiple stations and cross-station collaborative optimization of process parameters. This effectively overcomes the technical limitations of traditional single-station test calibration methods that ignore the thermal coupling effect between stations, and avoids defects such as incomplete fusion or over-fusion caused by the direct use of single-station cold-state parameters in subsequent stations due to shifts in the initial temperature field of the substrate. This significantly improves the consistency of multi-station laser welding processes and the quality of weld formation. Simultaneously, by introducing defocusing as an active adjustment variable and constraining its feasible region with dual physical boundaries of effective focus depth and spot coverage integrity, coupled with a soft penalty evaluation mechanism based on a dynamic process database, the optimized process parameters always remain within the calibrated and verified feasible process region. This avoids the risk of response surface model extrapolation and empowers the system with... The invention offers the ability to progressively explore unknown operating conditions and adaptively expand boundaries. Furthermore, by using pre-welding online temperature measurement data as the criterion for residual gating instead of directly correcting the input, it effectively suppresses the interference of welding spatter, fumes, and surface condition changes on the sensing signals, ensuring the robustness of thermal state perception and parameter correction. In addition, by deploying high-fidelity thermal accumulation solutions on the server side and degraded model calls on the production line's industrial PC side in a layered architecture, along with online learning of the equipment's recent response characteristics and dynamic updates of transition time margins, it unifies offline physical accuracy with online computational real-time performance, enabling complex algorithms to run stably within production line cycle time constraints. Therefore, this invention comprehensively improves the process adaptability, process stability, and product quality consistency of multi-station laser welding production lines under varying operating conditions, providing reliable process optimization and control technologies for high-quality welding manufacturing of large metal sheet parts such as automotive bodies.

Claims

1. A method for optimizing process parameters in high-speed laser welding of multi-station metal sheet parts, characterized in that, include: The station flow information, spatial location information, and heat dissipation characteristics information of each station fixture of the same workpiece in the multi-station laser welding production line are obtained. The process of transferring residual heat from the welding of the previous station to the welding area of ​​the next station is segmented and corrected to generate a cross-station thermal state offset map. Based on the cross-station thermal state offset map and the weld quality target of each subsequent station, the residual thermal state of the preceding station is converted into a correction coefficient for the laser thermal input of the current station. Based on the correction coefficient, the parameter combinations of each station are collaboratively mapped in the laser process parameter space, and a cross-station collaborative parameter scheme is obtained with the goal of minimizing the weld formation quality deviation. Based on the cross-station collaborative parameter scheme, production line cycle time constraints, and equipment response capabilities, the switching sequence of parameters for each station is dynamically arranged, and an executable process parameter optimization sequence is output.

2. The method for optimizing process parameters of high-speed laser welding of multi-station metal sheet parts according to claim 1, characterized in that, The steps for generating the cross-station thermal state migration map include: A thermal accumulation digital dictionary containing multiple working condition state vectors is pre-established. On the industrial side of the production line, the matching unit or interpolation unit of the dictionary is called according to the current working condition, and the residual heat temperature rise field across workstations is output using a reduced-order model. The working condition state vectors include at least plate characteristics, joint characteristics, fixture status, transfer time, and temperature status.

3. The method for optimizing process parameters of high-speed laser welding of multi-station metal sheet metal parts according to claim 2, characterized in that, The reduced-order model is used to output the residual heat rise field across workstations, including determining the residual heat rise field of the area to be welded in the subsequent workstation as follows: ； in, Indicates the location of the area to be welded in the subsequent workstation j. The equivalent temperature rise at the location; This represents the nth reduced-order basis function extracted from a high-fidelity snapshot on the server side; Represents the working condition state vector The corresponding expansion coefficients; This is the overall correction factor for the online temperature rise field; The correction coefficient Update according to the following residual gating method: ； ； in, This represents the candidate correction value for the k-th batch; These are the residual gating coefficients; This represents the residual between the pre-welding temperature value and the model prediction value for the current batch; This represents the variance of the residuals within the sliding window; The stable scale is determined by the sensor noise level and the historical residual noise band.

4. The method for optimizing process parameters of high-speed laser welding of multi-station metal sheet parts according to claim 3, characterized in that, The residual thermal state of the preceding station is converted into a correction coefficient for the laser heat input of the current station. The part that can effectively participate in the formation of the molten pool is screened from the residual temperature rise field through a spatial sensitivity function, and combined with temperature-related thermophysical properties, it is converted into residual heat energy. Then, after slow feedback correction with the pre-welding temperature measurement residual as the gate, the correction coefficient that should be retained for the external laser heat input of the current station is calculated in reverse according to the proportion of the residual energy to the heat input required by the single station reference.

5. The method for optimizing process parameters of high-speed laser welding of multi-station metal sheet metal parts according to claim 4, characterized in that, The residual thermal energy is calculated as follows: ； in, The corrected residual heat energy for welding segment q at station j. These are the model correction coefficients. For the effective volume to be welded, For temperature-dependent equivalent volumetric heat capacity, The spatial sensitivity function includes a transverse distribution with the weld centerline as the axis, a segmented distribution along the weld length, an attenuation distribution along the plate thickness direction, and an overlap interface correction factor.

6. The method for optimizing process parameters of high-speed laser welding of multi-station metal sheet metal parts according to claim 5, characterized in that, The collaborative mapping takes the total heat input balance determined by the correction coefficient as a hard constraint. Within the feasible region jointly defined by the dual physical boundaries of decoking amount and the stable process window, the goal is to minimize the comprehensive deviation of melt depth and melt width. The collaborative optimization mapping of power, speed and decoking amount is performed through an evaluation function with soft penalty terms.

7. The method for optimizing process parameters of high-speed laser welding of multi-station metal sheet parts according to claim 6, characterized in that, The coordinated solution for the power, speed, and defocus amount satisfies the following thermal input balance relationship: ； in, For laser power, For welding speed, This is the defocus amount. The length of the welded section. The effective absorption coefficient of the laser. The effective energy coupling factor corresponds to the defocusing amount and the characteristics of the current operating condition. To correct residual heat energy, The effective laser thermal input is used as a reference; and the defocusing amount is simultaneously constrained by the effective focal depth operating boundary and the operating boundary of the integrity of the spot coverage in the overlapping area.

8. A process parameter optimization system for high-speed laser welding of multi-station metal sheet parts, characterized in that, include: The thermal state perception and map generation module is used to acquire the station flow information, spatial position information and heat dissipation characteristic information of each station fixture of the same workpiece in the multi-station laser welding production line. It performs segmented correction on the transfer process of residual heat from the welding of the previous station to the welding area of ​​the subsequent station and generates a cross-station thermal state offset map. The heat input correction coefficient calculation module is used to convert the residual heat state of the previous station into the correction coefficient of the laser heat input of the current station based on the cross-station heat state offset map and the weld quality target of each subsequent station. The parameter collaborative mapping module is used to collaboratively map the parameter combinations of each station in the laser process parameter space according to the correction coefficient, and obtain a cross-station collaborative parameter scheme with the goal of minimizing the weld formation quality deviation. The timing arrangement and control output module is used to dynamically arrange the switching timing of parameters of each workstation according to the cross-workstation collaborative parameter scheme, production line cycle time constraints and equipment response capabilities, and output an executable process parameter optimization sequence.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the process parameter optimization method for high-speed laser welding of multi-station metal sheet metal as described in any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the process parameter optimization method for high-speed laser welding of multi-station metal sheet metal as described in any one of claims 1 to 7.

Images