B2B-based welding process parameter determination method, system and electronic device

Abstract

The application discloses a B2B-based welding process parameter determination method and system and electronic equipment, and relates to the technical field of metal welding. The method comprises the following steps: obtaining a plurality of to-be-screened combinations; each to-be-screened combination comprises a plurality of welding process parameters with different numerical values to be screened; a corresponding predicted value of a welding result is obtained based on each to-be-screened combination and a welding result prediction model; the welding result prediction model is obtained by training a B2B model using a training set; the training set comprises a plurality of training welding process parameter combinations and corresponding actual values of welding results; the B2B model comprises Bagging and Boosting; based on the predicted value of each welding result and a preset welding process parameter requirement, a target welding process parameter is determined from each to-be-screened combination, so that a welding operation is performed. The application reduces experimental costs and resource waste, and improves the quality and efficiency of the welding process.

Description

Technical Field

[0001] This invention relates to the field of metal welding technology, and in particular to a B2B-based method, system, and electronic device for determining welding process parameters. Background Technology

[0002] Metal welding, as an important manufacturing process, is widely used in many industries, including automotive, aerospace, construction, and electronics manufacturing. The quality of welding and the selection of process parameters are crucial to ensuring the strength, durability, and performance of welded joints. Traditionally, the optimization of welding process parameters has relied on experience and experimentation. Welding engineers search for the optimal solution by conducting numerous practical experiments with different parameter combinations. However, this approach has several serious drawbacks. First, it leads to expensive experimental costs and wasted resources. Second, too many variable factors or inaccurate finite element simulations can cause experimental failures. Furthermore, traditional methods typically require a large amount of experimental data to obtain results, which can be time-consuming. To overcome these problems, several data-driven and machine learning-based methods have been proposed in recent years to predict the optimal combination of welding process parameters. However, existing methods still face challenges, such as handling complex nonlinear relationships, noisy data, and high-dimensional features. Summary of the Invention

[0003] The purpose of this invention is to provide a method, system, and electronic device for determining welding process parameters based on B2B, thereby reducing experimental costs and resource waste, and improving the quality and efficiency of welding processes.

[0004] To achieve the above objectives, the present invention provides the following solution:

[0005] A B2B-based method for determining welding process parameters, comprising:

[0006] Multiple combinations to be screened are obtained; each combination to be screened includes multiple welding process parameters with different values ​​to be screened; the welding process parameters include: maximum temperature, holding time and cooling rate;

[0007] Based on each selection combination and the welding result prediction model, the predicted values ​​of the corresponding welding results are obtained. The welding result prediction model is obtained by training the B2B model using a training set. The training set includes multiple combinations of welding process parameters for training and the actual values ​​of the corresponding welding results. The B2B model includes Bagging and Boosting. The welding results include the strength of the weld and the residual stress after welding.

[0008] Based on the predicted values ​​of each welding result and the preset welding process parameter requirements, the target welding process parameters are determined from each of the selected combinations, and then the welding operation is performed.

[0009] Optionally, the training process of the welding result prediction model includes:

[0010] Collect the training set;

[0011] Based on the training set, multiple strong learners are constructed using the Bagging and Boosting algorithms; wherein the process of determining any one of the strong learners includes:

[0012] Based on the training set and sampling with replacement, multiple sample sets are obtained;

[0013] Determine the weights of each of the aforementioned sample sets;

[0014] Multiple weak learners are trained using sample sets with different weights to obtain trained weak learners.

[0015] The weights of each trained weak learner are determined based on their prediction performance.

[0016] Based on all trained weak learners and their corresponding weights, construct a strong learner;

[0017] The welding result prediction model is determined based on all of the strong learners.

[0018] Optionally, based on the training set and sampling with replacement, multiple sample sets are obtained, including:

[0019] The actual values ​​of each combination of welding process parameters used for training and the corresponding welding results in the training set are preprocessed to obtain a preprocessed training set; the preprocessing includes: data cleaning, data transformation and data feature dimensionality enhancement;

[0020] The data in the preprocessed training set are sampled with replacement to obtain multiple sample sets.

[0021] Optionally, the welding result prediction model is deployed on a cloud server or built locally.

[0022] Optionally, based on each combination to be screened and the welding result prediction model, the predicted values ​​of the corresponding welding results are obtained, including:

[0023] The welding process parameters of each combination to be screened are preprocessed to obtain the preprocessed combination to be screened; the preprocessing includes: data cleaning, data transformation and data feature dimensionality enhancement.

[0024] Each of the preprocessed combinations to be screened is input into the welding result prediction model to obtain the corresponding predicted value of the welding result.

[0025] A B2B-based welding process parameter determination system, comprising:

[0026] The module for obtaining combinations to be screened is used to obtain multiple combinations to be screened; each combination to be screened includes multiple welding process parameters with different values ​​to be screened; the welding process parameters include: maximum temperature, holding time and cooling rate;

[0027] The prediction module is used to obtain the predicted value of the corresponding welding result based on each combination to be screened and the welding result prediction model. The welding result prediction model is obtained by training the B2B model using a training set. The training set includes multiple combinations of welding process parameters for training and the actual values ​​of the corresponding welding results. The B2B model includes Bagging and Boosting. The welding results include the strength of the weld and the residual stress after welding.

[0028] The screening module is used to determine the target welding process parameters from each of the combinations to be screened based on the predicted values ​​of each welding result and the preset welding process parameter requirements, so as to perform the welding operation.

[0029] An electronic device includes a memory and a processor, the memory storing a computer program, and the processor running the computer program to cause the electronic device to perform the B2B-based welding process parameter determination method described above.

[0030] Optionally, the memory is a readable storage medium.

[0031] According to specific embodiments provided by the present invention, the present invention discloses the following technical effects:

[0032] This invention discloses a B2B-based method, system, and electronic device for determining welding process parameters. First, multiple selection combinations are obtained; each selection combination includes multiple welding process parameters with different values. Then, predicted welding results are obtained for each selection combination and a welding result prediction model. The welding result prediction model is trained using a training set on a B2B model, which includes multiple training combinations of welding process parameters and their corresponding actual welding result values. The B2B model includes Bagging and Boosting, and the welding results include weld strength and post-weld residual stress. Finally, based on the predicted values ​​of each welding result and preset welding process parameter requirements, target welding process parameters are determined from each selection combination, thereby performing the welding operation. This invention first uses the trained welding result prediction model to predict the welding results for each selection combination, then determines the optimal welding process parameters based on the preset welding process parameter requirements and the predicted values. This eliminates the need for actual welding experiments to determine the optimal welding process parameters, reducing experimental costs and resource waste, and improving the quality and efficiency of the welding process. Attached Figure Description

[0033] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0034] Figure 1 This is a schematic diagram of the B2B-based welding process parameter determination method provided in Embodiment 1 of the present invention;

[0035] Figure 2 This is a flowchart for finding the optimal combination of welding process parameters based on a B2B algorithm.

[0036] Figure 3 A flowchart of the B2B algorithm training process for finding the optimal combination of welding process parameters;

[0037] Figure 4 A schematic diagram of the training network structure for the B2B algorithm to find the optimal combination of welding process parameters;

[0038] Figure 5 The graph shows the evaluation and prediction results of the model for finding the optimal combination of welding process parameters based on the B2B algorithm.

[0039] Figure 6 This is a schematic diagram of the B2B algorithm prediction process for finding the optimal combination of welding process parameters. Detailed Implementation

[0040] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0041] The purpose of this invention is to provide a method, system, and electronic device for determining welding process parameters based on B2B, aiming to reduce experimental costs and resource waste, and improve the quality and efficiency of welding processes.

[0042] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0043] Example 1

[0044] Figure 1This is a schematic flowchart of the B2B-based welding process parameter determination method provided in Embodiment 1 of the present invention. Figure 1 As shown, the B2B-based welding process parameter determination method in this embodiment includes:

[0045] Step 101: Obtain multiple combinations to be filtered.

[0046] Each selection combination includes multiple welding process parameters with different values; the welding process parameters include: maximum temperature, holding time, and cooling rate.

[0047] Step 102: Based on each combination to be screened and the welding result prediction model, obtain the corresponding predicted value of the welding result.

[0048] The welding result prediction model is obtained by training the B2B model using a training set. The training set includes multiple combinations of welding process parameters for training and the corresponding actual values ​​of welding results. The B2B model includes Bagging and Boosting. The welding results include the weld strength and post-weld residual stress.

[0049] As an optional implementation, step 102 includes:

[0050] The welding process parameters of each combination to be screened are preprocessed to obtain the preprocessed combination to be screened; the preprocessing includes: data cleaning, data transformation and data feature dimensionality enhancement.

[0051] Each preprocessed combination to be screened is input into the welding result prediction model to obtain the corresponding predicted value of the welding result.

[0052] As an optional implementation method, the training process of the welding result prediction model includes:

[0053] Collect training data.

[0054] Based on the training set, multiple strong learners are constructed using the Bagging and Boosting algorithms; the process of determining any one of the strong learners includes:

[0055] Multiple sample sets are obtained based on the training set and sampling with replacement.

[0056] Determine the weights for each sample set.

[0057] Multiple weak learners are trained using sample sets with different weights to obtain trained weak learners.

[0058] The weights of each trained weak learner are determined based on their prediction performance.

[0059] A strong learner is constructed based on all trained weak learners and their corresponding weights.

[0060] A welding result prediction model is determined based on all strong learners.

[0061] As an optional implementation, multiple sample sets are obtained based on the training set and sampling with replacement, including:

[0062] The actual values ​​of each combination of welding process parameters used in the training set and the corresponding welding results are preprocessed to obtain the preprocessed training set. The preprocessing includes data cleaning, data transformation and data feature dimensionality enhancement.

[0063] The data in the preprocessed training set are sampled with replacement to obtain multiple sample sets.

[0064] As an alternative implementation, the welding result prediction model can be deployed on a cloud server or built locally.

[0065] Step 103: Based on the predicted values ​​of each welding result and the preset welding process parameter requirements, determine the target welding process parameters from each combination to be screened, and then perform the welding operation.

[0066] To further illustrate Example 1, the following are also provided: Figure 2 The flowchart shown is for finding the optimal combination of welding process parameters based on the B2B algorithm. The specific process is as follows:

[0067] Step S1: Collect welding experiment data.

[0068] Welding, as an important manufacturing process, plays a crucial role in various materials and applications. To optimize the welding process and ensure the strength of the weld joint and minimize residual stress, it is necessary to determine the optimal combination of welding process parameters. Step S1 aims to collect welding experimental data, including welding process parameters and corresponding welding results. The collection of experimental data is the foundation for building predictive models; therefore, it needs to broadly cover different welding parameters and actual welding results. Welding process parameters include, but are not limited to, maximum temperature, holding time, and cooling rate, as these parameters directly affect the quality and performance of the weld joint. Welding results include, but are not limited to, weld strength and post-weld residual stress.

[0069] Welding experimental data are mainly collected through high-temperature vacuum brazing experiments in a vacuum brazing furnace. In the experiment, experimental schemes can be set for different welding parameters, a series of welding experiments can be carried out, and the welding results can be recorded.

[0070] In summary, the goal of step S1 is to establish a dataset containing actual values ​​of welding process parameters and corresponding welding results, providing a foundation for subsequent model training and optimization. This dataset is the core basis for the prediction model, helping to determine the optimal combination of welding process parameters, thereby improving the quality and performance of welded joints.

[0071] Step S2: Experimental data preprocessing.

[0072] In predicting the optimal combination of welding process parameters, data preprocessing is a crucial step. The goal of this step is to ensure that the dataset is clean, accurate, and in a consistent format for subsequent model training and analysis. The preprocessing procedure is as follows:

[0073] First, perform data cleaning. Data cleaning is a crucial step in ensuring data quality. It involves detecting and correcting errors, outliers, or missing values ​​in the data. In welding experiments, issues such as measuring equipment malfunctions, human error in data entry, or other data corruption may occur. Data cleaning helps to eliminate these problems. Specific operations include:

[0074] (1) Detect and delete duplicate data records.

[0075] (2) Identify and process missing data. Missing values ​​can be filled by interpolation, deletion or substitution.

[0076] (3) Detecting and handling outliers: Statistical methods or domain knowledge can be used to identify outliers and decide whether to delete or correct them.

[0077] Second, data transformation. Data transformation involves adapting data to a form suitable for model training. This may include operations such as standardization, normalization, or logarithmic transformation to ensure consistency in numerical ranges across different features and that the data satisfies model assumptions. In the case of welding parameters, this might involve mapping parameter values ​​to a specific numerical range so that the model can better understand and process them.

[0078] The formula for data standardization is:

[0079]

[0080] Where z is the standardized value; x is the original data point value; μ is the mean of the dataset; and σ is the standard deviation of the dataset. Standardization is a data processing method used to transform data into a standard normal distribution with a specific mean and standard deviation (mean = 0, standard deviation = 1). Standardization helps eliminate proportional differences between different features, making the data easier for machine learning algorithms to process.

[0081] Third, feature enhancement. Since only a few welding process parameters are involved in the experiment, this may lead to underfitting of the experimental results. Therefore, it is necessary to perform feature enhancement on the collected experimental data to expand the number of features. The specific feature enhancement formula is as follows:

[0082] X = Select(x1+x2+x3+x4+...x n ) 2 (2).

[0083] Where x1, x2, x3, x4, ... x n All represent welding process parameters involved in the experiment, such as temperature, cooling rate, holding time, etc. Select indicates the selection operation, used to select features with only two items, and X represents the data that needs to be upgraded after selection.

[0084] Unlike previous methods, this invention proposes for the first time to perform feature dimensionality upscaling on experimental data to prevent underfitting. In addition, a selection operation is performed on the dimensionality-upgraded data, retaining only the data of the second degree terms, in order to prevent the model from overfitting if too much dimensionality-upgraded data is used.

[0085] Step S3: B2B model training.

[0086] like Figure 3 and Figure 4 As shown, the B2B algorithm, which combines the features of Bagging and Boosting, is used in the optimal combination prediction of welding process parameters. The B2B algorithm is an ensemble learning method that combines the advantages of Bagging and Boosting to improve the stability and accuracy of the model. The Bagging algorithm trains multiple weak learners multiple times through random sampling, and then averages the predictions of each weak learner to reduce the model's variance. The Boosting algorithm focuses on progressively training multiple weak learners, with each learner attempting to correct the errors of the previous learner to reduce model bias. The B2B algorithm combines these two methods, iteratively training multiple learners, with each learner trained based on the performance of the previous learner, thus obtaining a more powerful model.

[0087] Step S3, the intent recognition model, includes the following sub-steps:

[0088] Step A1: Data input.

[0089] The data preprocessed in step S2 is input into the training model. Through continuous iteration of the model, accurate prediction results can eventually be achieved.

[0090] Step A2: Sampling with replacement.

[0091] This invention proposes a sampling with replacement operation on the data to obtain multiple sets of samples. Because the cost and time required for a single welding experiment are very high, it is impossible to conduct hundreds of experiments for simulation. Therefore, this invention proposes sampling with replacement on limited data, allowing the limited data to form multiple different sets of data, and then using these data for model training.

[0092] By using sampling with replacement, on the one hand, experimental time is greatly reduced and training speed is increased. Model training with limited data allows for the rapid identification of the optimal combination of process parameters for brazing experiments. On the other hand, it reduces resource waste. The more brazing experiments conducted, the more resources are consumed; however, sampling with replacement allows a small number of experiments to achieve the same results as a large number of experiments, significantly reducing resource waste.

[0093] Step A3: Determine the sample weights.

[0094] In each iteration, a new learner is generated on the training set. This learner is then used to predict all samples to evaluate the importance of each sample. Specifically, the algorithm assigns a weight to each sample. Each time, the trained learner predicts samples; if a sample is predicted more accurately, its weight is decreased; otherwise, its weight is increased. Samples with higher weights have greater weight in the next training iteration, meaning that samples that are harder to distinguish become more important during training.

[0095] The formula for calculating sample weights is as follows:

[0096]

[0097] Among them, W n-1 W represents the weight of the (n-1)th sample. n α represents the weight of the nth sample. m G represents the weights of the learner, y represents the true value, and G represents the weights of the learner. m (x) represents the predicted value.

[0098] Step A4: Construct a weak learner.

[0099] In step A4, weak learners are constructed by inputting the processed data into the model. Each weak learner is trained on the data using a prediction formula to obtain preliminary prediction results.

[0100] The specific prediction formula for each weak learner is as follows:

[0101] G m (x)=W(W1X1+W2X2+W3X3+...+W n X n ) ⑷

[0102] Among them, W1, W2, W3 and W n X1, X2, X3, and X represent the weights of each sample, respectively. n The parameters represent the welding process parameters for each sample; W represents the matrix used to adjust the dimensions of all feature results; G m (x) represents the prediction result of the weak learner.

[0103] Unlike previous methods, the weak learner prediction proposed in this invention assigns weights to each feature, increasing the weight of features that are predicted incorrectly in each round of learning and decreasing the weight of features that are predicted correctly. This allows the learner to focus more on information that is predicted incorrectly in each round, ultimately achieving correct predictions for all results.

[0104] Step A5: Determine the learner weights.

[0105] This step assigns appropriate weights to each weak learner, increasing the weight of learners with good prediction performance and decreasing the weight of learners with poor prediction performance, ultimately training a strong learner that can make accurate predictions.

[0106] The formula for calculating the weights of a weak learner is as follows:

[0107] ε m =P(G m (x)≠y) (5).

[0108]

[0109] Where y represents the original data result; ε m α represents the error rate, which is the probability that the predicted result is not equal to the actual result. m The weights of the weak learners are represented by ln; the logarithmic function is represented by ln; and the m-th learner is represented by m.

[0110] Step A6: Determine N strong learners.

[0111] After steps A1-A5, a strong learner with predictive capabilities is obtained. The specific formula for the strong learner is as follows:

[0112]

[0113] Where G(X) represents a strong learner and M represents the total number of weak learners.

[0114] This invention uses an e-exponential function to process the final result of the strong learner calculation formula to prevent negative values. After step A6, a strong learner is obtained. However, the predictive effect of a single learner is limited. Therefore, this invention proposes to construct N strong learners and use the average of these N prediction results to obtain the final prediction result.

[0115] Step S4: Model evaluation.

[0116] In predicting the optimal combination of welding process parameters, evaluating model performance is a crucial step in ensuring model effectiveness. Using selected evaluation metrics to measure the model's performance on the test set, the evaluation results can be interpreted, and appropriate measures can be taken to improve the model. Step S4 comprehensively evaluates the model's performance, determining its effectiveness in predicting the optimal combination of welding process parameters. This helps ensure that the model produces reliable results in practical applications, thereby improving the quality and efficiency of the welding process.

[0117] The model evaluation formula is as follows:

[0118]

[0119] Here, Eval represents the evaluation result. Indicates the prediction result, y i This represents the actual result. The accuracy of the model's predictions is evaluated by calculating the difference between the actual and predicted values.

[0120] This invention proposes to use the e-exponential function to evaluate the prediction model. On the one hand, it can amplify the results for data with a large gap between the true value and the predicted value, and can be used to consider whether to retrain the model. On the other hand, the e-exponential function can prevent the occurrence of negative values.

[0121] Calculations show that, Figure 5 As shown, the model of this invention achieves a prediction accuracy of 91%, demonstrating high predictive performance.

[0122] Step S5: Model Deployment.

[0123] Once the model is built, the trained model will be deployed. The deployment options include deployment to a cloud server or local setup. Once deployment is complete, predictions can be made.

[0124] Step S6: Cross-experiment prediction.

[0125] like Figure 6As shown, the model is used for cross-experiment prediction to verify its performance in practical applications. This invention generates 100,000 sets of samples randomly, given a certain threshold for each process parameter, and combines different welding parameters. The model records the results of each cross-experiment, including the actual parameter combination used and the actual welding result. Finally, the brazing process parameter combination with the highest welding strength is obtained through screening.

[0126] For example, the maximum temperature of vacuum brazing is generally between 1100℃ and 1200℃, the holding time generally varies from 10 minutes to 60 minutes, and the filler metal thickness is generally 30µm, 60µm, 90µm, and 120µm. This invention uses random numbers to generate 200 sets of data from the 1100℃-1200℃ range and 100 sets of data from the 10-minute-60-minute holding time range. These three processes are then combined and the model of this invention is executed accordingly. Model verification shows that when the maximum temperature is 1150℃, the holding time is 25 minutes, and the filler metal thickness is 120µm, the welding strength reaches its maximum of 357 MPa.

[0127] In summary, the model of this invention can adapt to different process and material conditions through continuous learning. Based on the model's recommendations, welding processes can be optimized to improve the quality and performance of welded joints. The model's predictions can help reduce experimental costs and resource waste, and improve the efficiency of process parameters.

[0128] Example 2

[0129] The B2B-based welding process parameter determination system in this embodiment includes:

[0130] The module for acquiring combinations to be screened is used to acquire multiple combinations to be screened. Each combination to be screened includes multiple welding process parameters with different values. The welding process parameters include: maximum temperature, holding time, and cooling rate.

[0131] The prediction module is used to obtain the predicted value of the corresponding welding result based on each combination to be screened and the welding result prediction model. The welding result prediction model is obtained by training the B2B model using a training set. The training set includes multiple combinations of welding process parameters for training and the actual values ​​of the corresponding welding results. The B2B model includes Bagging and Boosting. The welding results include the strength of the weld and the residual stress after welding.

[0132] The screening module is used to determine the target welding process parameters from each screening combination based on the predicted values ​​of each welding result and the preset welding process parameter requirements, so as to carry out the welding operation.

[0133] Example 3

[0134] An electronic device includes a memory and a processor. The memory stores a computer program, and the processor runs the computer program to cause the electronic device to perform the B2B-based welding process parameter determination method of Embodiment 1.

[0135] As an optional implementation, the memory is a readable storage medium.

[0136] The beneficial effects of this invention are:

[0137] 1. Data processing.

[0138] This invention makes adjustments to data processing. Traditional data processing methods mainly involve normalization, standardization, data filling and deletion. This invention, for the first time, proposes to expand data features by using a specific formula to expand the original small amount of features into more features, thereby improving the underfitting problem of the model and further improving the model's prediction accuracy.

[0139] 2. Sampling with replacement.

[0140] This invention employs sampling with replacement for the collected data, eliminating the need for large-scale experiments and multiple verifications. It typically requires less experimental data to obtain accurate prediction results, thereby reducing the cost of data collection and experimentation.

[0141] 3. Highly accurate predictions.

[0142] The B2B algorithm combines the advantages of Bagging and Boosting algorithms, building a strong learner model by training multiple learners and iterating repeatedly. This method can effectively handle complex nonlinear relationships and noisy data, thus providing highly accurate predictions of welding process parameter combinations, far superior to traditional methods.

[0143] 4. High stability and robustness.

[0144] Because B2B algorithms possess the robustness of Bagging and Boosting algorithms, the model is better able to handle various complex real-world scenarios, including noisy data and variations in data distribution. This improves the model's stability, enabling it to perform well in different processes and environments.

[0145] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the systems disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the descriptions are relatively simple; relevant parts can be referred to the method section.

[0146] This document uses specific examples to illustrate the principles and implementation methods of the present invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of the present invention. Furthermore, those skilled in the art will recognize that, based on the ideas of the present invention, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A method for determining welding process parameters based on B2B, characterized in that, The method includes: Multiple combinations to be screened are obtained; each combination to be screened includes multiple welding process parameters with different values ​​to be screened; the welding process parameters include: maximum temperature, holding time and cooling rate; Based on each selection combination and the welding result prediction model, the predicted values ​​of the corresponding welding results are obtained. The welding result prediction model is obtained by training the B2B model using a training set. The training set includes multiple combinations of welding process parameters for training and the actual values ​​of the corresponding welding results. The B2B model includes Bagging and Boosting. The welding results include the strength of the weld and the residual stress after welding. Based on the predicted values ​​of each welding result and the preset welding process parameter requirements, the target welding process parameters are determined from each of the combinations to be screened, and then the welding operation is performed. The training process of the welding result prediction model includes: Collect the training set; Based on the training set, multiple strong learners are constructed using the Bagging and Boosting algorithms; wherein the process of determining any one of the strong learners includes: Based on the training set and sampling with replacement, multiple sample sets are obtained; Determine the weights of each of the aforementioned sample sets; Multiple weak learners are trained using sample sets with different weights to obtain trained weak learners. The weights of each trained weak learner are determined based on their prediction performance. Based on all trained weak learners and their corresponding weights, construct a strong learner; The welding result prediction model is determined based on all of the strong learners; Based on each selection combination and the welding result prediction model, the predicted values ​​of the corresponding welding results are obtained, including: The welding process parameters of each combination to be screened are preprocessed to obtain the preprocessed combination to be screened; the preprocessing includes: data cleaning, data transformation and data feature dimensionality enhancement. Each of the preprocessed combinations to be screened is input into the welding result prediction model to obtain the corresponding predicted value of the welding result.

2. The method for determining welding process parameters based on B2B according to claim 1, characterized in that, Based on the training set and sampling with replacement, multiple sample sets are obtained, including: The actual values ​​of each combination of welding process parameters used for training and the corresponding welding results in the training set are preprocessed to obtain a preprocessed training set; the preprocessing includes: data cleaning, data transformation and data feature dimensionality enhancement; The data in the preprocessed training set are sampled with replacement to obtain multiple sample sets.

3. The method for determining welding process parameters based on B2B according to claim 1, characterized in that, The welding result prediction model is deployed on a cloud server or built locally.

4. A B2B-based welding process parameter determination system, characterized in that, The system includes: The module for obtaining the combination to be screened is used to obtain multiple combinations to be screened; each combination to be screened includes multiple welding process parameters with different values ​​to be screened; the welding process parameters include: maximum temperature, holding time and cooling rate; The prediction module is used to obtain the predicted value of the corresponding welding result based on each combination to be screened and the welding result prediction model. The welding result prediction model is obtained by training the B2B model using a training set. The training set includes multiple combinations of welding process parameters for training and the actual values ​​of the corresponding welding results. The B2B model includes Bagging and Boosting. The welding results include the strength of the weld and the residual stress after welding. The screening module is used to determine the target welding process parameters from each of the combinations to be screened based on the predicted values ​​of each welding result and the preset welding process parameter requirements, so as to perform the welding operation. The training process of the welding result prediction model includes: Collect the training set; Based on the training set, multiple strong learners are constructed using the Bagging and Boosting algorithms; wherein the process of determining any one of the strong learners includes: Based on the training set and sampling with replacement, multiple sample sets are obtained; Determine the weights of each of the aforementioned sample sets; Multiple weak learners are trained using sample sets with different weights to obtain trained weak learners. The weights of each trained weak learner are determined based on their prediction performance. Based on all trained weak learners and their corresponding weights, construct a strong learner; The welding result prediction model is determined based on all of the strong learners; Based on each selection combination and the welding result prediction model, the predicted values ​​of the corresponding welding results are obtained, including: The welding process parameters of each combination to be screened are preprocessed to obtain the preprocessed combination to be screened; the preprocessing includes: data cleaning, data transformation and data feature dimensionality enhancement. Each of the preprocessed combinations to be screened is input into the welding result prediction model to obtain the corresponding predicted value of the welding result.

5. An electronic device, characterized in that, The device includes a memory and a processor, the memory being used to store a computer program, and the processor running the computer program to cause the electronic device to perform the B2B-based welding process parameter determination method according to any one of claims 1 to 3.

6. An electronic device according to claim 5, characterized in that, The memory is a readable storage medium.

Images