Welding parameter adaptive matching method and system for multi-material welding

Abstract

This invention provides an adaptive matching method and system for welding parameters in multi-material welding, belonging to the field of welding process control technology. The method includes: acquiring welding monitoring data of the previous batch of base materials during the welding process, and calculating the batch thermal response deviation rate and welding power output deviation rate; calculating the net thermal response deviation of the base material based on the batch thermal response deviation rate and welding power output deviation rate; inputting the net thermal response deviation of the base material, the grade code of the previous batch of base materials, and the grade code of the next batch of base materials into a pre-constructed batch characteristic migration evaluation model to obtain the expected thermal response deviation evaluation value for the next batch; determining the correction direction and correction magnitude of the initial welding parameters for the next batch of base materials based on the expected thermal response deviation evaluation value; and matching and correcting the initial welding parameters for the next batch of base materials according to the correction direction and correction magnitude. This invention solves the problem of unstable welding quality caused by fluctuations in thermal response characteristics between different batches of base materials.

Description

Technical Field

[0001] This invention relates to the field of welding process control technology, and more specifically to an adaptive matching method and system for welding parameters for multi-material welding. Background Technology

[0002] In the welding of multi-material components, different base materials have different thermophysical properties such as thermal conductivity, specific heat capacity, and melting point. It is necessary to set corresponding welding parameters for different material combinations to ensure welding quality.

[0003] Currently, welding parameter settings mainly rely on process experiments and empirical data, and these parameters are solidified for batch production after offline optimization. However, different batches of the same grade of base material may have different thermal response characteristics due to factors such as slight fluctuations in raw material composition and differences in heat treatment conditions. Using the original parameters may lead to defects such as incomplete fusion and over-penetration, while retesting batch by batch would increase costs and extend the cycle time, making it difficult to meet the needs of flexible manufacturing.

[0004] Therefore, how to quickly compensate for the differences in thermal response between batches of base materials and avoid welding parameter mismatch is a problem that needs to be solved in welding process control. Summary of the Invention

[0005] This invention provides an adaptive matching method and system for welding parameters for multi-material welding, which is used to solve the problem of unstable welding quality caused by fluctuations in thermal response characteristics between different batches of base materials.

[0006] In a first aspect, the present invention provides an adaptive matching method for welding parameters for multi-material welding, comprising: Obtain welding monitoring data of the previous batch of base materials during the welding process, and calculate the batch thermal response deviation rate and welding power output deviation rate of the previous batch of base materials; Based on the batch thermal response deviation rate and the welding power output deviation rate, calculate the net thermal response deviation of the previous batch of base material; Input the net thermal response deviation of the base material, the grade code of the previous batch of base material and the grade code of the next batch of base material into the pre-constructed batch characteristic migration assessment model to obtain the expected thermal response deviation assessment value of the next batch output by the batch characteristic migration assessment model. Based on the expected thermal response deviation assessment value of the next batch, determine the correction direction and correction range of the initial welding parameters of the next batch of base material; Based on the correction direction and correction magnitude, the initial welding parameters of the next batch of base materials are matched and corrected.

[0007] Secondly, the present invention provides an adaptive matching system for welding parameters for multi-material welding, comprising: The monitoring data acquisition module is used to acquire the welding monitoring data of the previous batch of base materials during the welding process, and to calculate the batch thermal response deviation rate and welding power output deviation rate of the previous batch of base materials. The deviation rate calculation module is used to calculate the net thermal response deviation of the previous batch of base material based on the batch thermal response deviation rate and the welding power output deviation rate. The net thermal response calculation module is used to input the net thermal response deviation of the parent material, the grade code of the previous batch of parent material and the grade code of the next batch of parent material into a pre-constructed batch characteristic migration assessment model, and obtain the expected thermal response deviation assessment value of the next batch output by the batch characteristic migration assessment model. The migration assessment module is used to determine the correction direction and correction magnitude of the initial welding parameters of the next batch of base material based on the expected thermal response deviation assessment value of the next batch. The parameter correction module is used to match and correct the initial welding parameters of the next batch of base materials according to the correction direction and correction magnitude.

[0008] One or more technical solutions provided in this invention have at least the following technical effects or advantages: The technical solution of this invention first obtains the welding monitoring data of the previous batch of base materials during the welding process, calculates the batch thermal response deviation rate and welding power output deviation rate, and separates and quantifies the welding response changes caused by material composition fluctuations and thermophysical property differences of the same batch of base materials from the influence caused by the output fluctuations of the welding power source itself, so as to provide accurate and decoupled deviation input for the adaptive correction of welding parameters across batches.

[0009] Furthermore, based on the batch thermal response deviation rate and welding power output deviation rate, the net thermal response deviation of the previous batch of base material is calculated. The influence of welding power output fluctuation on weld width variation is separated from the overall deviation, and a pure deviation index that only reflects the material characteristics difference of the base material itself is obtained. This provides a reliable basis for accurately assessing the influence of material migration between batches and formulating targeted parameter correction strategies.

[0010] Furthermore, the net thermal response deviation of the base material, the grade code of the previous batch of base material, and the grade code of the next batch of base material are input into a pre-constructed batch characteristic migration assessment model to obtain the expected thermal response deviation assessment value of the next batch output by the model. By utilizing the deep learning capability of the model for material characteristic differences and migration patterns between grades, the material thermal response deviation exposed in the previous batch of production, combined with grade change information and batch drift trend, is transformed into a quantitative prediction of the welding thermal response performance of the next batch of base material.

[0011] Furthermore, based on the expected thermal response deviation assessment value of the next batch, the correction direction and correction magnitude of the initial welding parameters of the next batch of base material are determined. The predicted thermal response deviation value output by the model is transformed into a welding parameter adjustment instruction with directional and magnitude constraints. While making up for the difference in thermal response between batches, the risk of over-adjustment caused by the uncertainty of model prediction is suppressed by confidence weighting.

[0012] Finally, based on the correction direction and magnitude, the initial welding parameters of the next batch of base materials are matched and corrected. The adjustment strategy based on material migration assessment and confidence constraints is implemented into specific welding parameter settings, so that the next batch of base materials adopts current parameters that are compatible with its expected thermal response characteristics when welding starts, thus compensating for the weld width deviation that may be caused by material differences between batches from the source.

[0013] In summary, the technical solution of this invention achieves adaptive pre-adjustment of welding parameters across batches and grades, effectively compensating for differences in material thermal response between batches. While avoiding welding defects caused by batch fluctuations or grade switching, it also takes into account the pertinence and stability of the correction, thereby improving the consistency and pass rate of multi-material welding production. Attached Figure Description

[0014] Figure 1 This is a flowchart illustrating the adaptive matching method for welding parameters for multi-material welding provided by the present invention.

[0015] Figure 2 This is a schematic diagram of the batch characteristic migration evaluation model in the adaptive matching method for welding parameters for multi-material welding provided by the present invention.

[0016] Figure 3 This is a schematic diagram of the adaptive matching system for welding parameters for multi-material welding provided by the present invention.

[0017] In the attached diagram, the labels representing each component are as follows: The system includes a monitoring data acquisition module 11, a deviation rate calculation module 12, a net thermal response calculation module 13, a migration assessment module 14, and a parameter correction module 15. Detailed Implementation

[0018] This invention provides an adaptive matching method and system for welding parameters for multi-material welding, which solves the problem of unstable welding quality caused by fluctuations in thermal response characteristics between different batches of base materials.

[0019] It should be noted that the terms "comprising" and "having" are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or server that includes a series of steps or units is not necessarily limited to those steps or units that are explicitly listed, but may include other steps or modules that are not explicitly listed or that are inherent to these processes, methods, products, or devices.

[0020] Example 1, as Figure 1 As shown, this invention provides an adaptive matching method for welding parameters for multi-material welding, the method comprising: S100: Obtain welding monitoring data of the previous batch of base materials during the welding process, and calculate the batch thermal response deviation rate and welding power output deviation rate of the previous batch of base materials.

[0021] This step involves obtaining the actual weld width data and the actual power output data of the welding power source from the previous batch of base material welding process, and comparing them with the reference weld width value and the set power value to calculate the batch thermal response deviation rate and welding power output deviation rate.

[0022] Step S100 in the method provided by the present invention includes: Obtain the actual weld width value of the weld seam of the previous batch of base materials, and obtain the reference weld width value of the corresponding grade of the previous batch of base materials under the same set welding parameters from the grade welding parameter reference library; Calculate the deviation between the actual melt width value and the reference melt width value, and use it as the batch thermal response deviation rate; Obtain the measured average power of the welding power source and the set power value of the welding torch during the welding process of the previous batch of base materials; The deviation between the measured average power and the set power value is calculated and used as the welding power output deviation rate.

[0023] In this step, the actual weld width value of the previous batch of base metal welds is first obtained. This value can be directly obtained using a vision sensor integrated into the welding system. For example, a laser weld tracking sensor can be used to scan and extract the weld surface in real time after welding, and the actual weld width value can be calculated using an image processing algorithm. Simultaneously, the reference weld width value for the corresponding base metal grade under the same welding parameters is retrieved from the grade welding parameter reference library.

[0024] Among them, the welding parameter reference library is a pre-generated database. In a laboratory or stable production environment, standard samples of base materials of various grades that meet the standards in terms of composition and heat treatment are selected. System process tests are carried out for each grade under different combinations of welding speed, welding current and other parameters. The reference weld width, reference weld depth and other characteristic values ​​corresponding to each parameter combination are collected and stored, so that they can be called up later by base material grade and set welding parameters as query conditions.

[0025] For example, the base material of the previous batch was 6061 aluminum alloy, with a welding speed of 10 mm / s and a welding current of 120 A. Measurements of the weld seam after this batch were taken, and the actual weld width was found to be 2.2 mm. From the welding parameter reference library, based on the grade 6061 aluminum alloy and the set parameters of 10 mm / s welding speed and 120 A welding current, the corresponding reference weld width value was found to be 2.0 mm.

[0026] Secondly, calculate the deviation between the actual weld width and the reference weld width. The deviation is calculated as (actual weld width - reference weld width) / reference weld width. The result is used as the batch thermal response deviation rate. For example, based on the obtained actual weld width of 2.2 mm and the reference weld width of 2.0 mm, the batch thermal response deviation rate is calculated as (2.2 - 2.0) / 2.0 = 0.1, or 10%.

[0027] Next, obtain the measured average power of the welding power source and the set power value of the welding torch during the welding process of the previous batch of base materials. The above-mentioned measured average power and set power values ​​can be directly obtained through the voltage and current sampling module built into the welding power source.

[0028] For example, during the welding of the same batch of 6061 aluminum alloy, the instantaneous power data of the entire welding period was collected by the welding power source monitoring module, and the calculated average power was 2352W. Meanwhile, the welding torch was set to a power of 2400W before welding.

[0029] Finally, the deviation between the measured average power and the welding torch set power value is calculated. Deviation = (measured average power - welding torch set power value) / welding torch set power value. The calculation result is used as the welding power output deviation rate.

[0030] For example, based on the measured average power of 2352W and the welding torch setting power of 2400W, the welding power output deviation rate is calculated to be (2352-2400) / 2400=-0.02, or -2%.

[0031] It should be noted that the above values ​​are for illustrative purposes only and do not constitute a limitation on the present invention.

[0032] In summary, this step separates and quantifies the changes in welding response caused by fluctuations in material composition and differences in thermophysical properties of the same batch of base materials from the effects of fluctuations in the output of the welding power source itself. This provides accurate and decoupled deviation inputs for the adaptive correction of welding parameters across batches.

[0033] S200: Based on the batch thermal response deviation rate and the welding power output deviation rate, calculate the net thermal response deviation of the previous batch of base material.

[0034] This step, based on the batch thermal response deviation rate and welding power output deviation rate, uses the power-weld width response coefficient and linear approximation correction factor to derive the power-attributed weld width deviation rate from the measured average power, and then subtracts it from the batch thermal response deviation rate to obtain the net thermal response deviation of the base material.

[0035] Step S200 in the method provided by the present invention includes: The measured average power of the previous batch of base materials during the welding process is obtained, and the power-weld width response coefficient and linear approximation correction factor corresponding to the grade of the previous batch of base materials are obtained from the grade welding parameter reference library. Multiply the measured average power by the power-weld width response coefficient to obtain the theoretical weld width value; Multiply the theoretical melt width value by the linear approximation correction factor to obtain the corrected theoretical melt width value; Calculate the deviation between the corrected theoretical weld width value and the reference weld width value to obtain the power-attributed weld width deviation rate; The difference between the batch thermal response deviation rate and the power-attributed melt width deviation rate is taken as the net thermal response deviation of the base material.

[0036] In this step, the measured average power of the previous batch of base materials during the welding process is first obtained. Simultaneously, the power-weld width response coefficient and linear approximation correction factor for the corresponding base material grade are retrieved from the grade welding parameter reference library.

[0037] The power-weld width response coefficient and the linear approximation correction factor were both calibrated through standard process experiments and stored in the base material grade welding parameter reference library. The power-weld width response coefficient represents the change in weld width caused by a unit change in welding power for a given base material grade. It is obtained by measuring weld width while changing the welding power under fixed welding parameters, and then linearly fitting the power and weld width data to obtain the slope. The linear approximation correction factor is used to compensate for deviations in the power-weld width linear relationship under non-ideal conditions. It is obtained by calculating the ratio of the actual weld width value to the linearly estimated weld width value at each power point and taking the average. After calibration, these two parameters are stored in the base material grade welding parameter reference library for direct querying and retrieval by grade in subsequent steps.

[0038] For example, the base material of the previous batch was 6061 aluminum alloy. The measured average power was 2352W, which was obtained by collecting and calculating the voltage and current sampling module built into the welding power source. From the welding parameter reference library, based on the grade 6061 aluminum alloy, the power-weld width response coefficient of this grade was found to be 0.0008mm / W, and the linear approximation correction factor was 1.05.

[0039] Secondly, the measured average power is multiplied by the power-weld width response coefficient to obtain the theoretical weld width value. For example, by multiplying the obtained measured average power by the power-weld width response coefficient, the theoretical weld width value is calculated to be 2352 × 0.0008 = 1.8816 mm.

[0040] Next, multiply the theoretical weld width value by the linear approximation correction factor to obtain the corrected theoretical weld width value. For example, multiplying the calculated theoretical weld width value by the linear approximation correction factor yields a corrected theoretical weld width value of 1.8816 × 1.05 ≈ 1.9757 mm.

[0041] Next, the deviation between the corrected theoretical weld width and the reference weld width is calculated. The deviation is calculated as (corrected theoretical weld width - reference weld width) / reference weld width. The result is used as the power-attributed weld width deviation rate. For example, a reference weld width of 2.0 mm is obtained from the welding parameter reference library. Based on the obtained corrected theoretical weld width of 1.9757 mm, the power-attributed weld width deviation rate is calculated as (1.9757 - 2.0) / 2.0 = -0.01215, or approximately -1.22%.

[0042] Finally, the difference between the batch thermal response deviation rate and the power-attributed melt width deviation rate is taken as the net thermal response deviation of the base material. For example, the batch thermal response deviation rate has been calculated to be 10% in the previous steps. Based on the calculated power-attributed melt width deviation rate of -1.22%, the net thermal response deviation of the base material is calculated as 10% - (-1.22%) = 11.22%.

[0043] It should be noted that the above values ​​are for illustrative purposes only and do not constitute a limitation on the present invention.

[0044] In summary, this step separates the impact of welding power supply output fluctuations on weld width variation from the overall deviation, obtaining a pure deviation index that only reflects the differences in the material properties of the base material itself. This provides a reliable basis for subsequent accurate assessment of the impact of material migration between batches and for developing targeted parameter correction strategies.

[0045] S300: Input the net thermal response deviation of the base material, the grade code of the previous batch of base material and the grade code of the next batch of base material into the pre-constructed batch characteristic migration assessment model, and obtain the expected thermal response deviation assessment value of the next batch output by the batch characteristic migration assessment model.

[0046] like Figure 2 As shown, this step inputs the net thermal response deviation and grade code of the previous batch of base material and the grade code of the next batch of base material into a pre-built batch characteristic migration assessment model. By utilizing the model's deep learning capabilities on material characteristic differences and migration patterns between grades, it outputs the expected thermal response deviation assessment value for the next batch. The effect is to transform the material thermal response deviation exposed in the previous batch of production, combined with grade change information and batch drift trend, into a quantitative prediction of the welding thermal response performance of the next batch of base material.

[0047] Step S300 in the method provided by the present invention includes: Obtain the continuous batch drift trend of the previous batch of base material, wherein the continuous batch drift trend is the absolute value of the linear regression slope of the batch thermal response deviation rate sequence of the corresponding grade of the previous batch of base material in the most recent consecutive batches divided by the standard deviation of the batch thermal response deviation rate sequence. The net thermal response deviation of the base material, the drift trend of the continuous batch, the grade code of the previous batch of base material, and the grade code of the next batch of base material are input into a pre-constructed batch characteristic migration assessment model to obtain the expected thermal response deviation assessment value of the next batch output by the batch characteristic migration assessment model.

[0048] In this step, the continuous batch drift trend of the previous batch of base material is first obtained. The continuous batch drift trend is the absolute value of the linear regression slope of the thermal response deviation rate sequence of the corresponding grade of the previous batch of base material in the most recent consecutive batches, divided by the standard deviation of the thermal response deviation rate sequence of that batch.

[0049] For example, the previous batch of base material corresponded to the grade 6061 aluminum alloy. The batch thermal response deviation rate sequence for this grade in the most recent five consecutive batches was [2.1%, 3.5%, 5.8%, 6.2%, 8.4%]. A linear regression fit was performed on this sequence, yielding a slope of 1.58% / batch. Taking the absolute value of 1.58% / batch, the standard deviation of this sequence was calculated to be 2.21%. Therefore, the consecutive batch drift trend degree is approximately 1.58% / 2.21% ≈ 0.715.

[0050] Furthermore, the net thermal response deviation of the base material, the drift trend of consecutive batches, the grade code of the previous batch of base material, and the grade code of the next batch of base material are input into the pre-constructed batch characteristic migration assessment model to obtain the expected thermal response deviation assessment value of the next batch output by the model.

[0051] In this step, inputting the net thermal response deviation of the base material, the continuous batch drift trend, the grade code of the previous batch of base material, and the grade code of the next batch of base material into the pre-constructed batch characteristic migration evaluation model further includes: When the grade code of the next batch of base material is different from that of the previous batch of base material, the historical net thermal response deviation of the most recent historical batch of the grade code of the next batch of base material is used as the net thermal response deviation of the base material and input into the batch characteristic migration evaluation model.

[0052] Specifically, it determines whether the grade code of the next batch of parent material is the same as that of the previous batch: If the grade code of the next batch of parent material is different from that of the previous batch, the net thermal response deviation of the previous batch of parent material is no longer used as input. Instead, the net thermal response deviation of the next batch of parent material recorded in the most recent historical batch is retrieved from the historical database and used as the net thermal response deviation input item, replacing the original net thermal response deviation of the previous batch of parent material. This is then input into the batch characteristic migration assessment model along with other input items. If the grade code of the next batch of parent material is the same as that of the previous batch of parent material, the net thermal response deviation of the previous batch of parent material is directly used as input and processed according to the conventional procedure.

[0053] For example, the previous batch's base material grade was 6061 aluminum alloy, with a net thermal response deviation of 11.22% and a continuous batch drift trend of 0.715. The next batch's base material grade is 304 stainless steel, which is different from the previous batch. The net thermal response deviation of the most recent historical batch of 304 stainless steel is retrieved from the historical database, yielding a value of -3.8%. This -3.8% is used as the input for the net thermal response deviation, along with the continuous batch drift trend of 0.715, the previous batch's base material grade code AL6061, and the next batch's base material grade code SS304, and input into a pre-built batch characteristic migration assessment model to obtain the expected thermal response deviation assessment value for the next batch.

[0054] In this step, the pre-construction of the batch characteristic migration evaluation model includes: Collect multiple sets of historical welding batch data. Each set of historical welding batch data includes the net thermal response deviation of the base material of a historical batch, the drift trend of consecutive batches, the grade code of the current batch and the grade code of the next batch, as well as the actual batch thermal response deviation rate of the next batch. A batch characteristic migration evaluation model is constructed, wherein the batch characteristic migration evaluation model includes a grade embedding layer, a deviation feature encoding layer, and an evaluation output layer; The grade embedding layer is used to map the grade code of the current batch and the grade code of the next batch into grade feature vectors respectively. The deviation feature encoding layer is used to encode the net thermal response deviation of the base material and the drift trend degree of the continuous batches, and output the deviation feature vector. The evaluation output layer is used to concatenate the grade feature vector of the current batch, the grade feature vector of the next batch and the deviation feature vector, and evaluate and output the expected thermal response deviation evaluation value of the next batch. The batch characteristic migration evaluation model is trained using the historical welding batch data until it converges, resulting in a pre-constructed batch characteristic migration evaluation model.

[0055] Specifically, multiple sets of historical welding batch data are first collected. Each set of data includes the net thermal response deviation of the base material for a historical batch, the drift trend of consecutive batches, the grade code of the current batch and the grade code of the next batch, as well as the actual batch thermal response deviation rate of the next batch.

[0056] For example, one set of historical data collected is as follows: the current batch is 6061 aluminum alloy, coded AL6061, with a net thermal response deviation of 8.5% and a continuous batch drift trend of 0.62; the next batch is also 6061 aluminum alloy, coded AL6061, with an actual batch thermal response deviation rate of 9.8%. Another set of cross-grade historical data is as follows: the current batch is 6061 aluminum alloy, coded AL6061, with a net thermal response deviation of 11.2% and a continuous batch drift trend of 0.71; the next batch is 304 stainless steel, coded SS304, with an actual batch thermal response deviation rate of -4.1%.

[0057] Secondly, a batch feature transfer evaluation model is constructed, which includes a brand embedding layer, a bias feature encoding layer, and an evaluation output layer. The brand embedding layer is a trainable embedding matrix that maps each unique brand code to a fixed-dimensional brand feature vector. The mapping process is a lookup operation, using the input brand code as an index to retrieve the corresponding vector from the embedding matrix.

[0058] In this design, the embedding vector has a fixed dimension, meaning it does not change with the number of grade types. The standard for this fixed dimension is to provide sufficient vector space for different grades to express material property differences in most application scenarios, while keeping the number of model parameters within a reasonable range. For example, an empirical value of 128 dimensions can be used as the embedding vector dimension. When there are 30 grade types, the embedding matrix size is 30×128; when the number of grade types increases to 100, the embedding matrix size expands accordingly to 100×128, while the embedding dimension remains unchanged at 128. This fixed-dimensional setting method is simple to implement and avoids the problem of frequent model structure adjustments due to changes in grade types.

[0059] For example, the collected historical data involves 30 different brands, and the brand embedding layer is set to 128 dimensions. The brand code AL6061 is input into the brand embedding layer, and after lookup table mapping via the embedding matrix, a 128-dimensional brand feature vector is output, with some values ​​in the range [0.12, -0.34, 0.56, 0.08, -0.19]. The brand code SS304 is input into the same embedding layer, and after lookup table mapping, another 128-dimensional brand feature vector is output, with some values ​​in the range [-0.21, 0.43, -0.18, 0.33, 0.07].

[0060] Next, a deviation feature encoding layer is constructed, consisting of one or more fully connected layers. This layer encodes the two scalar values—the net thermal response deviation of the parent material and the drift trend of consecutive batches—and outputs a deviation feature vector of fixed dimensions. The weight matrix and bias vector of the fully connected layer are determined through training, and the activation function can be the ReLU nonlinear function.

[0061] For example, the deviation feature encoding layer is set as a two-layer fully connected network. The first layer expands the 2-dimensional input to 32-dimensional, and the second layer maps the 32-dimensional to 64-dimensional output. The net thermal response deviation of the parent material of 8.5% and the continuous batch drift trend degree of 0.62 are combined into a 2-dimensional vector [0.085, 0.62] and input into this layer. After the first layer weight matrix is ​​calculated and ReLU is activated, and then after the second layer weight matrix is ​​calculated and ReLU is activated, a 64-dimensional deviation feature vector is output, in which some values ​​are [0.05, 0.31, 0.00, 0.22, 0.14].

[0062] Next, an evaluation output layer is constructed. First, the feature vectors of the current batch and the next batch output from the brand embedding layer are concatenated with the deviation feature vector output from the deviation feature encoding layer to obtain a comprehensive feature vector. This concatenated comprehensive feature vector is then passed through one or more fully connected layers, ultimately outputting a scalar value as the expected thermal response deviation evaluation value for the next batch. The last fully connected layer in the output layer has no activation function and directly outputs a linear result, allowing the model to output any real value as the expected thermal response deviation evaluation value for the next batch.

[0063] For example, the 128-dimensional feature vector of the current batch, the 128-dimensional feature vector of the next batch, and the 64-dimensional deviation feature vector are concatenated into a 320-dimensional comprehensive feature vector. This comprehensive feature vector is then fed into two fully connected layers. The first layer maps the 320 dimensions to 64 dimensions and applies ReLU activation, while the second layer maps the 64 dimensions to a 1-dimensional scalar output, resulting in an expected thermal response deviation assessment value of 9.5% for the next batch.

[0064] Finally, the model was trained using multiple sets of historical welding batch data. The actual batch thermal response deviation rate of the next batch in each set of data was used as the ground truth label, and the model's output of the expected thermal response deviation assessment value for the next batch was used as the predicted value. A loss function, such as the mean squared error loss function, was constructed. A gradient descent optimization algorithm was used, and the weight parameters of each layer were updated through backpropagation. Iterative training continued until the loss function value converged to the preset conditions, resulting in a pre-constructed batch characteristic transfer assessment model.

[0065] Among them, the preset convergence condition can adopt an early stopping strategy, that is, monitor the loss value of the model on the validation set, and when the loss value has not been lower than the historical minimum loss value in 10 consecutive iterations, it is determined that the model has converged and the training is terminated.

[0066] For example, historical data is divided into training and validation sets in a 4:1 ratio. Mean squared error is used as the loss function, and the Adam optimizer is employed with an initial learning rate of 0.001 and a batch size of 32. Multiple iterations are performed on the training set, calculating the loss and updating the model parameters in reverse at each iteration. Loss changes are monitored on the validation set; training stops when the validation set loss fails to decrease for 10 consecutive iterations, and the model parameters with the minimum validation set loss are saved as a pre-built batch feature transfer evaluation model.

[0067] For example, continuing the aforementioned scenario, if the previous batch's base material was 6061 aluminum alloy, the calculated net thermal response deviation was 11.22%, and the continuous batch drift trend was 0.715. The next batch's base material is 304 stainless steel. Inputting the base material net thermal response deviation of 11.22%, the continuous batch drift trend of 0.715, the previous batch grade code AL6061, and the next batch grade code SS304 into a pre-built batch characteristic migration assessment model, the output is an expected thermal response deviation assessment value of -5.8% for the next batch.

[0068] This step also includes: Based on the root mean square error of the prediction of the batch characteristic migration assessment model on the historical validation set, the assessment confidence is calculated and obtained. The assessment confidence is negatively correlated with the root mean square error of the prediction. The historical validation set is obtained based on the historical welding batch data.

[0069] Specifically, a portion of the collected historical welding batch data that was not used for model training is set as a historical validation set. For example, the portion of historical welding batch data used for model training is removed, and the remaining data is used as the historical validation set, which does not overlap with the training set.

[0070] Furthermore, the net thermal response deviation of the parent material, the continuous batch drift trend, the current batch grade code, and the next batch grade code of each data set in the historical validation set are input into the trained and converged batch characteristic transfer evaluation model to obtain the expected thermal response deviation evaluation value for the next batch for each group. The evaluation value of each group is compared with the actual batch thermal response deviation rate of the corresponding next batch to calculate the root mean square error of prediction. The root mean square error of prediction is calculated as follows: first, the square of the difference between the evaluation value and the actual value of each group is calculated; then, the mean of all squared values ​​is calculated; finally, the square root of the mean is taken.

[0071] For example, validation set data is input into the model group by group to obtain evaluation values. Taking one group as an example, the model outputs an evaluation value of 9.5%, while the actual batch thermal response deviation rate for this group is 9.8%, and the error for this group is (9.5% - 9.8%). 2 =0.000009. Summarizing all the squared errors, the mean is 0.000072. Taking the square root, the root mean square error of the prediction is approximately 0.085%.

[0072] Finally, the assessment confidence level is calculated based on the root mean square error of the prediction, and the assessment confidence level is negatively correlated with the root mean square error of the prediction. Specifically, the root mean square error of the prediction is divided by the historical standard deviation of the batch thermal response deviation rate, the quotient is increased by 1, and the reciprocal is taken to obtain the assessment confidence level, that is, assessment confidence level = 1 / (1 + root mean square error of prediction / historical standard deviation of batch thermal response deviation rate).

[0073] Among them, the historical standard deviation of the batch thermal response deviation rate is the standard deviation of the actual batch thermal response deviation rate of the next batch corresponding to each group of data in the historical validation set, which is used to normalize the root mean square error of the prediction.

[0074] For example, by dividing the obtained root mean square error of prediction (RMSE) of 0.085% by the historical standard deviation of the batch thermal response deviation rate (1.7%), the confidence level of the assessment is calculated as 1 / (1+0.085% / 1.7%)≈0.9524, or 95.24%.

[0075] It should be noted that the above values ​​are for illustrative purposes only and do not constitute a limitation on the present invention.

[0076] In summary, this step provides a direct numerical basis for the pre-adjustment of subsequent welding parameters, avoiding welding quality defects caused by material batch fluctuations or grade changes.

[0077] S400: Based on the expected thermal response deviation assessment value of the next batch, determine the correction direction and correction magnitude of the initial welding parameters of the next batch of base material.

[0078] This step determines the correction direction based on the positive or negative sign of the expected thermal response deviation assessment value for the next batch, and calculates the correction magnitude by combining its absolute value with the correction magnitude benchmark coefficient and assessment confidence level.

[0079] Step S400 in the method provided by the present invention includes: When the expected thermal response deviation assessment value for the next batch is positive, the correction direction is determined to be to reduce the initial welding current; when the expected thermal response deviation assessment value for the next batch is negative, the correction direction is determined to be to increase the initial welding current; when the expected thermal response deviation assessment value for the next batch is zero, the correction direction is determined to be to maintain the initial welding current unchanged. The absolute value of the expected thermal response deviation assessment value for the next batch is multiplied by a preset correction range benchmark coefficient, and the resulting product is used as the theoretical correction range. The theoretical correction range is then multiplied by the assessment confidence level to obtain the correction range coefficient.

[0080] In this step, the correction direction is first determined based on the sign of the expected thermal response deviation assessment value for the next batch. When the assessment value is positive, the correction direction is to reduce the initial welding current; when the assessment value is negative, the correction direction is to increase the initial welding current; and when the assessment value is zero, the correction direction is to keep the initial welding current unchanged.

[0081] For example, if the base material for the next batch is 304 stainless steel, the expected thermal response deviation for the next batch output by the batch characteristic migration assessment model is -5.8%. Since -5.8% is a negative value, the correction direction is determined to be to increase the initial welding current.

[0082] Furthermore, the absolute value of the expected thermal response deviation assessment value for the next batch is multiplied by a preset correction range benchmark coefficient, and the resulting product is used as the theoretical correction range. The preset correction range benchmark coefficient needs to be calibrated based on the response characteristics of the welding process system. The calibration method is as follows: During the process debugging phase, for typical base materials, a known change in welding current is artificially introduced, and the corresponding change ratio of weld width is observed to establish a mapping relationship between the current change rate and the weld width change rate, thereby determining the correction range benchmark coefficient. The correction range benchmark coefficient is usually set between 0.5 and 2.0. If the weld width is highly sensitive to current changes, a smaller value, such as 0.5, can be used; if the sensitivity is low, a larger value, such as 1.5 or 2.0, can be used.

[0083] For example, the correction margin base coefficient is set to 1.5, which means that when the expected thermal response deviation is 1%, the theoretical correction margin is a current adjustment of 1.5%. Taking the absolute value of the evaluation value as 5.8%, the theoretical correction margin is calculated to be 5.8% × 1.5 = 8.7%.

[0084] Finally, the theoretical correction magnitude is multiplied by the assessment confidence level to obtain the correction magnitude coefficient. For example, if the assessment confidence level has been calculated as 95.24% in the previous step, multiplying the theoretical correction magnitude of 8.7% by the assessment confidence level of 95.24% yields a correction magnitude coefficient of 8.7% × 0.9524 ≈ 8.29%.

[0085] It should be noted that the above values ​​are for illustrative purposes only and do not constitute a limitation on the present invention.

[0086] In summary, this step transforms the predicted thermal response deviation from the model into welding parameter adjustment instructions with directionality and amplitude constraints, thereby pre-correcting the initial welding current of the next batch of base materials. While compensating for the differences in thermal response between batches, the confidence-weighted approach suppresses the risk of over-adjustment that may be caused by the uncertainty of model predictions, thus balancing the pertinence and stability of the correction.

[0087] S500: Based on the correction direction and correction magnitude, the initial welding parameters of the next batch of base materials are matched and corrected.

[0088] This step involves matching and correcting the initial welding current value of the next batch of base material retrieved from the grade welding parameter reference library, based on the determined correction direction and correction magnitude coefficient.

[0089] Step S500 in the method provided by the present invention includes: Obtain the initial welding current value of the corresponding grade of the next batch of base material from the grade welding parameter reference library; The initial welding current value is corrected using the correction amplitude coefficient, and the initial welding parameters of the next batch of base material are matched and corrected according to the correction amplitude.

[0090] In this step, the initial welding current value for the next batch of base material is first obtained from the welding parameter reference library. For example, if the next batch of base material is 304 stainless steel, the initial welding current value for 304 stainless steel is retrieved from the welding parameter reference library, and the initial welding current value is found to be 200A.

[0091] Furthermore, a correction amplitude coefficient is used to correct the initial welding current value. Based on the determined correction direction, if the correction direction is to increase, the initial welding current value is multiplied by (1 + correction amplitude coefficient); if the correction direction is to decrease, the initial welding current value is multiplied by (1 - correction amplitude coefficient); if the correction direction is to maintain, the correction amplitude coefficient is zero, and the initial welding current value remains unchanged. The calculated result is the corrected welding current value, completing the matching correction.

[0092] For example, if the determined correction direction is to increase the initial welding current and the correction magnitude coefficient is 8.29%, then the calculated corrected welding current value is 200A×1.0829≈216.6A. The corrected welding current value will be used as the initial welding current setting value for the next batch of 304 stainless steel base material welding.

[0093] It should be noted that the above values ​​are for illustrative purposes only and do not constitute a limitation on the present invention.

[0094] In summary, this step translates the adjustment strategy derived from material migration assessment and confidence constraints into specific welding parameter settings. This ensures that the next batch of base material adopts current parameters that match its expected thermal response characteristics when welding begins, thereby compensating for potential weld width deviations caused by material differences between batches at the source and improving the consistency and pass rate of cross-batch welding.

[0095] In summary, this invention achieves adaptive pre-adjustment of welding parameters across batches and grades, effectively compensating for differences in material thermal response between batches. While avoiding welding defects caused by batch fluctuations or grade switching, it also takes into account the pertinence and stability of the correction, improving the consistency and pass rate of multi-material welding production.

[0096] Example 2, as Figure 3 As shown, this invention provides an adaptive matching system for welding parameters for multi-material welding, the system comprising: The monitoring data acquisition module 11 is used to acquire the welding monitoring data of the previous batch of base materials during the welding process, and to calculate the batch thermal response deviation rate and welding power output deviation rate of the previous batch of base materials.

[0097] The step of acquiring welding monitoring data of the previous batch of base materials during the welding process and calculating the batch thermal response deviation rate and welding power output deviation rate of the previous batch of base materials includes: Obtain the actual weld width value of the weld seam of the previous batch of base materials, and obtain the reference weld width value of the corresponding grade of the previous batch of base materials under the same set welding parameters from the grade welding parameter reference library; Calculate the deviation between the actual melt width value and the reference melt width value, and use it as the batch thermal response deviation rate; Obtain the measured average power of the welding power source and the set power value of the welding torch during the welding process of the previous batch of base materials; The deviation between the measured average power and the set power value is calculated and used as the welding power output deviation rate.

[0098] The deviation rate calculation module 12 is used to calculate the net thermal response deviation of the previous batch of base material based on the batch thermal response deviation rate and the welding power output deviation rate.

[0099] The step of calculating the net thermal response deviation of the previous batch of base material based on the batch thermal response deviation rate and the welding power output deviation rate includes: The measured average power of the previous batch of base materials during the welding process is obtained, and the power-weld width response coefficient and linear approximation correction factor corresponding to the grade of the previous batch of base materials are obtained from the grade welding parameter reference library. Multiply the measured average power by the power-weld width response coefficient to obtain the theoretical weld width value; Multiply the theoretical melt width value by the linear approximation correction factor to obtain the corrected theoretical melt width value; Calculate the deviation between the corrected theoretical weld width value and the reference weld width value to obtain the power-attributed weld width deviation rate; The difference between the batch thermal response deviation rate and the power-attributed melt width deviation rate is taken as the net thermal response deviation of the base material.

[0100] The net thermal response calculation module 13 is used to input the net thermal response deviation of the parent material, the grade code of the previous batch of parent material and the grade code of the next batch of parent material into a pre-constructed batch characteristic migration assessment model, and obtain the expected thermal response deviation assessment value of the next batch output by the batch characteristic migration assessment model.

[0101] The step of inputting the net thermal response deviation of the base material, the grade code of the previous batch of base material, and the grade code of the next batch of base material into a pre-constructed batch characteristic migration assessment model to obtain the expected thermal response deviation assessment value of the next batch output by the batch characteristic migration assessment model includes: Obtain the continuous batch drift trend of the previous batch of base material, wherein the continuous batch drift trend is the absolute value of the linear regression slope of the batch thermal response deviation rate sequence of the corresponding grade of the previous batch of base material in the most recent consecutive batches divided by the standard deviation of the batch thermal response deviation rate sequence. The net thermal response deviation of the base material, the drift trend of the continuous batch, the grade code of the previous batch of base material, and the grade code of the next batch of base material are input into a pre-constructed batch characteristic migration assessment model to obtain the expected thermal response deviation assessment value of the next batch output by the batch characteristic migration assessment model.

[0102] The step of inputting the net thermal response deviation of the base material, the continuous batch drift trend, the grade code of the previous batch of base material, and the grade code of the next batch of base material into a pre-constructed batch characteristic migration evaluation model further includes: When the grade code of the next batch of base material is different from that of the previous batch of base material, the historical net thermal response deviation of the most recent historical batch of the grade code of the next batch of base material is used as the net thermal response deviation of the base material and input into the batch characteristic migration evaluation model.

[0103] The pre-construction of the batch characteristic migration evaluation model includes: Collect multiple sets of historical welding batch data. Each set of historical welding batch data includes the net thermal response deviation of the base material of a historical batch, the drift trend of consecutive batches, the grade code of the current batch and the grade code of the next batch, as well as the actual batch thermal response deviation rate of the next batch. A batch characteristic migration evaluation model is constructed, wherein the batch characteristic migration evaluation model includes a grade embedding layer, a deviation feature encoding layer, and an evaluation output layer; The grade embedding layer is used to map the grade code of the current batch and the grade code of the next batch into grade feature vectors respectively. The deviation feature encoding layer is used to encode the net thermal response deviation of the base material and the drift trend degree of the continuous batches, and output the deviation feature vector. The evaluation output layer is used to concatenate the grade feature vector of the current batch, the grade feature vector of the next batch and the deviation feature vector, and evaluate and output the expected thermal response deviation evaluation value of the next batch. The batch characteristic migration evaluation model is trained using the historical welding batch data until it converges, resulting in a pre-constructed batch characteristic migration evaluation model.

[0104] This also includes: Based on the root mean square error of the prediction of the batch characteristic migration assessment model on the historical validation set, the assessment confidence is calculated and obtained. The assessment confidence is negatively correlated with the root mean square error of the prediction. The historical validation set is obtained based on the historical welding batch data.

[0105] The migration assessment module 14 is used to determine the correction direction and correction magnitude of the initial welding parameters of the next batch of base material based on the expected thermal response deviation assessment value of the next batch.

[0106] The step of determining the correction direction and magnitude of the initial welding parameters for the next batch of base material based on the expected thermal response deviation assessment value of the next batch includes: When the expected thermal response deviation assessment value for the next batch is positive, the correction direction is determined to be to reduce the initial welding current; when the expected thermal response deviation assessment value for the next batch is negative, the correction direction is determined to be to increase the initial welding current; when the expected thermal response deviation assessment value for the next batch is zero, the correction direction is determined to be to maintain the initial welding current unchanged. The absolute value of the expected thermal response deviation assessment value for the next batch is multiplied by a preset correction range benchmark coefficient, and the resulting product is used as the theoretical correction range. The theoretical correction range is then multiplied by the assessment confidence level to obtain the correction range coefficient.

[0107] The parameter correction module 15 is used to match and correct the initial welding parameters of the next batch of base materials according to the correction direction and correction magnitude.

[0108] The step of matching and correcting the initial welding parameters of the next batch of base materials according to the correction direction and correction magnitude includes: Obtain the initial welding current value of the corresponding grade of the next batch of base material from the grade welding parameter reference library; The initial welding current value is corrected using the correction amplitude coefficient, and the initial welding parameters of the next batch of base material are matched and corrected according to the correction amplitude.

[0109] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

[0110] This specification and accompanying drawings are merely illustrative examples of the invention and are intended to cover any and all modifications, variations, combinations, or equivalents within the scope of the invention. Clearly, those skilled in the art can make various alterations and modifications to the invention without departing from its scope. Therefore, if such modifications and modifications fall within the scope of the invention and its equivalents, the invention is intended to include these modifications and modifications.

Claims

1. An adaptive matching method for welding parameters for multi-material welding, characterized in that, include: Obtain welding monitoring data of the previous batch of base materials during the welding process, and calculate the batch thermal response deviation rate and welding power output deviation rate of the previous batch of base materials; Based on the batch thermal response deviation rate and the welding power output deviation rate, calculate the net thermal response deviation of the previous batch of base material; Input the net thermal response deviation of the base material, the grade code of the previous batch of base material and the grade code of the next batch of base material into the pre-constructed batch characteristic migration assessment model to obtain the expected thermal response deviation assessment value of the next batch output by the batch characteristic migration assessment model. Based on the expected thermal response deviation assessment value of the next batch, determine the correction direction and correction range of the initial welding parameters of the next batch of base material; Based on the correction direction and correction magnitude, the initial welding parameters of the next batch of base materials are matched and corrected.

2. The adaptive matching method for welding parameters for multi-material welding according to claim 1, characterized in that, The step of acquiring welding monitoring data from the previous batch of base materials during the welding process and calculating the batch thermal response deviation rate and welding power output deviation rate of the previous batch of base materials includes: Obtain the actual weld width value of the weld seam welded in the previous batch of base materials, and obtain the reference weld width value of the corresponding grade of the previous batch of base materials under the same set welding parameters from the grade welding parameter reference library; Calculate the deviation between the actual melt width value and the reference melt width value, and use it as the batch thermal response deviation rate; Obtain the measured average power of the welding power source and the set power value of the welding torch during the welding process of the previous batch of base materials; The deviation between the measured average power and the set power value is calculated and used as the welding power output deviation rate.

3. The adaptive matching method for welding parameters for multi-material welding according to claim 1, characterized in that, The calculation of the net thermal response deviation of the previous batch of base material based on the batch thermal response deviation rate and the welding power output deviation rate includes: The measured average power of the previous batch of base materials during the welding process is obtained, and the power-weld width response coefficient and linear approximation correction factor corresponding to the grade of the previous batch of base materials are obtained from the grade welding parameter reference library. Multiply the measured average power by the power-weld width response coefficient to obtain the theoretical weld width value; Multiply the theoretical melt width value by the linear approximation correction factor to obtain the corrected theoretical melt width value; Calculate the deviation between the corrected theoretical weld width value and the reference weld width value to obtain the power-attributed weld width deviation rate; The difference between the batch thermal response deviation rate and the power-attributed melt width deviation rate is taken as the net thermal response deviation of the base material.

4. The adaptive matching method for welding parameters for multi-material welding according to claim 1, characterized in that, The step of inputting the net thermal response deviation of the base material, the grade code of the previous batch of base material, and the grade code of the next batch of base material into a pre-constructed batch characteristic migration assessment model, and obtaining the expected thermal response deviation assessment value of the next batch output by the batch characteristic migration assessment model, includes: Obtain the continuous batch drift trend of the previous batch of base material, wherein the continuous batch drift trend is the absolute value of the linear regression slope of the batch thermal response deviation rate sequence of the corresponding grade of the previous batch of base material in the most recent consecutive batches divided by the standard deviation of the batch thermal response deviation rate sequence. The net thermal response deviation of the base material, the drift trend of the continuous batch, the grade code of the previous batch of base material, and the grade code of the next batch of base material are input into a pre-constructed batch characteristic migration assessment model to obtain the expected thermal response deviation assessment value of the next batch output by the batch characteristic migration assessment model.

5. The adaptive matching method for welding parameters for multi-material welding according to claim 4, characterized in that, The step of inputting the net thermal response deviation of the base material, the continuous batch drift trend, the grade code of the previous batch of base material, and the grade code of the next batch of base material into a pre-constructed batch characteristic migration evaluation model further includes: When the grade code of the next batch of base material is different from that of the previous batch of base material, the historical net thermal response deviation of the most recent historical batch of the grade code of the next batch of base material is used as the net thermal response deviation of the base material and input into the batch characteristic migration evaluation model.

6. The adaptive matching method for welding parameters for multi-material welding according to claim 4, characterized in that, The pre-construction of the batch characteristic migration evaluation model includes: Collect multiple sets of historical welding batch data. Each set of historical welding batch data includes the net thermal response deviation of the base material of a historical batch, the drift trend of consecutive batches, the grade code of the current batch and the grade code of the next batch, as well as the actual batch thermal response deviation rate of the next batch. A batch characteristic migration evaluation model is constructed, wherein the batch characteristic migration evaluation model includes a grade embedding layer, a deviation feature encoding layer, and an evaluation output layer; The grade embedding layer is used to map the grade code of the current batch and the grade code of the next batch into grade feature vectors respectively. The deviation feature encoding layer is used to encode the net thermal response deviation of the base material and the drift trend degree of the continuous batches, and output the deviation feature vector. The evaluation output layer is used to concatenate the grade feature vector of the current batch, the grade feature vector of the next batch and the deviation feature vector, and evaluate and output the expected thermal response deviation evaluation value of the next batch. The batch characteristic migration evaluation model is trained using the historical welding batch data until it converges, resulting in a pre-built batch characteristic migration evaluation model.

7. The adaptive matching method for welding parameters for multi-material welding according to claim 6, characterized in that, Also includes: Based on the root mean square error of the prediction of the batch characteristic migration assessment model on the historical validation set, the assessment confidence is calculated and obtained. The assessment confidence is negatively correlated with the root mean square error of the prediction. The historical validation set is obtained based on the historical welding batch data.

8. The adaptive matching method for welding parameters for multi-material welding according to claim 1, characterized in that, The determination of the correction direction and magnitude of the initial welding parameters for the next batch of base material based on the expected thermal response deviation assessment value of the next batch includes: When the expected thermal response deviation assessment value for the next batch is positive, the correction direction is determined to be to reduce the initial welding current; when the expected thermal response deviation assessment value for the next batch is negative, the correction direction is determined to be to increase the initial welding current; when the expected thermal response deviation assessment value for the next batch is zero, the correction direction is determined to be to maintain the initial welding current unchanged. The absolute value of the expected thermal response deviation assessment value for the next batch is multiplied by a preset correction range benchmark coefficient, and the resulting product is used as the theoretical correction range. The theoretical correction range is then multiplied by the assessment confidence level to obtain the correction range coefficient.

9. The adaptive matching method for welding parameters for multi-material welding according to claim 1, characterized in that, The step of matching and correcting the initial welding parameters of the next batch of base materials according to the correction direction and correction magnitude includes: Obtain the initial welding current value of the corresponding grade of the next batch of base material from the grade welding parameter reference library; The initial welding current value is corrected using the correction amplitude coefficient, and the initial welding parameters of the next batch of base material are matched and corrected according to the correction amplitude.

10. An adaptive matching system for welding parameters for multi-material welding, characterized in that, The adaptive matching method for welding parameters for multi-material welding as described in any one of claims 1 to 9 includes: The monitoring data acquisition module is used to acquire the welding monitoring data of the previous batch of base materials during the welding process, and to calculate the batch thermal response deviation rate and welding power output deviation rate of the previous batch of base materials. The deviation rate calculation module is used to calculate the net thermal response deviation of the previous batch of base material based on the batch thermal response deviation rate and the welding power output deviation rate. The net thermal response calculation module is used to input the net thermal response deviation of the parent material, the grade code of the previous batch of parent material and the grade code of the next batch of parent material into a pre-constructed batch characteristic migration assessment model, and obtain the expected thermal response deviation assessment value of the next batch output by the batch characteristic migration assessment model. The migration assessment module is used to determine the correction direction and correction magnitude of the initial welding parameters of the next batch of base material based on the expected thermal response deviation assessment value of the next batch. The parameter correction module is used to match and correct the initial welding parameters of the next batch of base materials according to the correction direction and correction magnitude.

Images