Welding parameter prediction method and apparatus, and electronic device and storage medium

Abstract

The embodiments of the present application relate to the technical field of welding. Disclosed are a welding parameter prediction method and apparatus, and an electronic device and a storage medium. The method comprises: acquiring a feature value of a first welding parameter at a first moment; and inputting the feature value of the first welding parameter into a pre-trained prediction model to obtain a feature value of a second welding parameter at the moment following the first moment, which is output by the prediction model, wherein the second welding parameter includes at least one of welding time and welding power, and the first welding parameter is an independent variable of the second welding parameter. The embodiments of the present application can accurately predict the welding time and welding power of a battery cell, thereby effectively improving the accuracy of a welding time and welding power prediction result; moreover, the embodiments can also intervene in a welding process on the basis of the prediction result in a timely manner, thereby improving the welding quality.

Description

Welding parameter prediction methods, devices, electronic equipment and storage media

[0001] This application claims priority to Chinese Patent Application No. 202510288841.5, filed on March 11, 2025, entitled “Method, Apparatus, Electronic Device and Storage Medium for Predicting Welding Parameters”, the disclosure of which is incorporated herein by reference in its entirety. Technical Field

[0002] This application relates to the field of welding technology, and more specifically, to a welding parameter prediction method, apparatus, electronic device, and storage medium. Background Technology

[0003] Ultrasonic welding of electrode tabs is a crucial step in battery cell production, and its quality directly affects the cell's performance, safety, and production costs. Welding time and welding power are two key factors in selecting welding parameters. Welding time refers to the duration of ultrasonic waves applied to the welding point. If the welding time is too short, the welding may be insufficient, resulting in a weak connection and affecting the overall performance of the battery cell. Conversely, excessive welding time may cause over-welding or burn-through, damaging the cell materials and affecting product quality and safety. Therefore, correctly setting the welding time is essential to ensuring welding quality. Simultaneously, welding power refers to the applied energy intensity. If the power is too low, the welding may be insufficient; if the power is too high, it may lead to over-welding or burn-through, causing excessive heating and damage to the materials.

[0004] In the traditional ultrasonic welding parameter selection process, key data such as welding time and welding power can only be obtained after actual welding experiments. Because welding time and welding power can only be obtained afterward, the identification and resolution of welding problems are delayed. Therefore, predicting welding time and welding power and intervening in a timely manner based on the prediction results to improve battery welding quality has become a top priority. Summary of the Invention

[0005] This application provides a welding parameter prediction method, apparatus, electronic device, and storage medium to solve the technical problem in related technologies where welding time and welding power can only be obtained after the fact, resulting in the delay in the identification and resolution of welding problems.

[0006] According to a first aspect of the embodiments of this application, a welding parameter prediction method is provided, the method comprising:

[0007] Obtain the feature value of the first welding parameter at the first moment; wherein, the first welding parameter includes at least one of the material parameter of the welding operation object and the welding process parameter;

[0008] The feature value of the first welding parameter is input into a pre-trained prediction model; wherein the prediction model is trained using the parameter values ​​of the first welding parameter at multiple sample times as training samples, through a graph convolutional network and a gated recurrent network; each training sample also includes a training label, the training label being the feature value of the second welding parameter at the next time of the corresponding sample time; wherein the second welding parameter includes at least one of welding time and welding power, and the first welding parameter is the independent variable of the second welding parameter;

[0009] The feature values ​​of the second welding parameters at the next time step after the first time step are obtained from the output of the prediction model.

[0010] As an optional implementation, the graph convolutional network includes a graph convolution module and an attention module; the gated recurrent network includes a feature extraction module and a second welding parameter prediction module.

[0011] The feature values ​​of the first welding parameter are input into a pre-trained prediction model to obtain the feature values ​​of the second welding parameter at the next time step after the first time step, as output by the prediction model, including:

[0012] The feature values ​​of the first welding parameter are input into the graph convolution module and the attention module respectively to obtain the target feature matrix output by the graph convolution module and the target feature vector output by the attention module; the updated feature values ​​of the first welding parameter are obtained based on the target feature matrix and the target feature vector.

[0013] The updated feature values ​​of the first welding parameters are input into the feature extraction module to obtain the target hidden state output by the feature extraction module. The target hidden state indicates the dependency relationship between the first welding parameters. The target hidden state is input into the second welding parameter prediction module to output the feature values ​​of the second welding parameters at the next moment.

[0014] As an optional implementation, the graph convolution module includes at least one graph convolution unit; there is a predetermined relationship between the first welding parameters;

[0015] The step of inputting the feature values ​​of the first welding parameter into the graph convolution module to obtain the target feature matrix output by the graph convolution module includes:

[0016] Based on the characteristic values ​​of the first welding parameters, a feature map is determined at the first moment; wherein, the nodes of the feature map are the first welding parameters, and the edges between the nodes are the predetermined relationships between the first welding parameters.

[0017] The feature maps are sequentially input into each graph convolutional unit to obtain the feature matrices output by each graph convolutional unit, and the feature matrix output by the last graph convolutional unit is used as the target feature matrix.

[0018] As an optional implementation, the attention module includes at least one attention unit and a global pooling unit; wherein each attention unit includes an attention channel corresponding to the first welding parameter;

[0019] The step of inputting the feature values ​​of the first welding parameter into the attention module to obtain the target feature vector output by the attention module includes:

[0020] Determine the input data for each attention unit; where, for the first attention unit, the input data is the feature value of the first welding parameter; for non-first attention units, the input data is the feature value output after convolution calculation of the previous attention unit;

[0021] For each attention unit, the input data corresponding to each first welding parameter is input into the corresponding attention channel, and convolution calculation is performed in the corresponding attention channel to obtain the feature value output by each attention channel; wherein, the weight value of each attention channel is different;

[0022] The feature values ​​output by each attention channel in the last attention unit are input into the global pooling unit. The maximum value among all feature values ​​in each attention channel is extracted as the corresponding feature value after updating the first welding parameters. The feature values ​​after updating the first welding parameters are combined into a one-dimensional feature vector, which is used as the target feature vector.

[0023] As an optional implementation, the feature extraction module includes multiple gated loop units and a fully connected unit;

[0024] The step of inputting the feature values ​​of the updated first welding parameters into the feature extraction module and obtaining the target hidden state output by the feature extraction module includes:

[0025] For the first gated loop unit, the input data is the feature value after the first welding parameter is updated, and the output data is the hidden state of the first gated loop unit;

[0026] For non-first gated loop units, the input data are the feature values ​​after the first welding parameter update and the hidden state output by the previous gated loop unit, and the output is the hidden state of the corresponding gated loop unit.

[0027] The hidden state output by the last gated loop unit is input to the fully connected unit for full connection processing, and the hidden state output by the fully connected unit is obtained as the target hidden state.

[0028] As an optional implementation, the second welding parameter prediction module includes multiple gated loop units and a global pooling unit;

[0029] The step of inputting the target hiding state into the second welding parameter prediction module and outputting the feature value of the second welding parameter at the next moment includes:

[0030] Determine the input data for each gated loop unit; wherein, for the first gated loop unit, the input data is the target hiding state; for non-first gated loop units, the input data is the hiding state output by the previous gated loop unit;

[0031] For each gated loop unit, perform a dot product operation on the hidden state of its own output and output the predicted value of the second welding parameter;

[0032] The predicted values ​​of the second welding parameters output by each gated loop unit are input into the global pooling unit. The predicted values ​​of all the second welding parameters are pooled to obtain the feature values ​​of the second welding parameters at the next time step.

[0033] As an optional implementation, the prediction model is generated in the following ways:

[0034] The parameter values ​​of the first welding parameter and the second welding parameter at multiple time points are obtained in advance. The parameter values ​​of the first welding parameter and the second welding parameter at each time point are preprocessed to obtain the feature values ​​of the first welding parameter and the second welding parameter at each time point. The preprocessing includes at least one of data cleaning and normalization.

[0035] Based on the characteristic values ​​of the first welding parameter and the second welding parameter at each time point, determine the characteristic values ​​of the first welding parameter at multiple sample times and the characteristic value of the second welding parameter at the next time point of the corresponding sample time.

[0036] The initial prediction model is trained using the first welding parameter feature value at the sample time as the training sample and the second welding parameter feature value at the next time of the corresponding sample time as the training label, until the training stop condition is met, thus obtaining the welding prediction model.

[0037] According to a second aspect of the embodiments of this application, a welding parameter prediction device is provided, the device comprising:

[0038] The first processing module is configured to acquire feature values ​​of a first welding parameter at a first moment; wherein the first welding parameter includes at least one of the material parameters of the object being welded and the welding process parameters.

[0039] The second processing module is configured to input the feature values ​​of the first welding parameter into a pre-trained prediction model; wherein the prediction model is trained using the parameter values ​​of the first welding parameter at multiple sample times as training samples, through a graph convolutional network and a gated recurrent network; each training sample further includes a training label, the training label being the feature value of the second welding parameter at the next time step of the corresponding sample time; wherein the second welding parameter includes at least one of welding time and welding power, and the first welding parameter is the independent variable of the second welding parameter;

[0040] The third processing module is configured to obtain the feature values ​​of the second welding parameters at the next time step after the first time step output by the prediction model.

[0041] According to a third aspect of the embodiments of this application, an electronic device is provided, including: a memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps of the method described in any of the first aspects.

[0042] According to a fourth aspect of the embodiments of this application, a computer-readable storage medium is provided having a computer program stored thereon that, when executed by a processor, implements the steps of the method as described in any of the first aspects.

[0043] The beneficial effects of the technical solutions provided in this application are:

[0044] In this embodiment, the parameter values ​​of the first welding parameter at multiple sample times are used as training samples, and the feature values ​​of the second welding parameter at the next time step of the corresponding sample time are used as training labels to train the graph convolutional network and gated recurrent units to obtain the prediction model. The feature values ​​of the first welding parameter at the first time step are obtained and input into the pre-trained prediction model to obtain the feature values ​​of the second welding parameter at the next time step of the first time step output by the prediction model. Compared to the traditional method of obtaining welding time and welding power through actual welding experiments, this embodiment can accurately predict the welding time and welding power of the battery cell, effectively improving the accuracy of the welding time and welding power prediction results, and instructing personnel to intervene in a timely manner based on the prediction results to improve welding quality. Attached Figure Description

[0045] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments of this application will be briefly introduced below.

[0046] Figure 1 is a flowchart illustrating a welding parameter prediction method provided in an embodiment of this application;

[0047] Figure 2 is a schematic diagram of the structure of a prediction model provided in an embodiment of this application;

[0048] Figure 3 is a schematic diagram of the structure of a graph convolution module provided in an embodiment of this application;

[0049] Figure 4 is a schematic diagram of an attention module provided in an embodiment of this application;

[0050] Figure 5 is a schematic diagram of a feature extraction module provided in an embodiment of this application;

[0051] Figure 6 is a schematic diagram of the structure of a second welding parameter prediction module provided in an embodiment of this application;

[0052] Figure 7 is a structural schematic diagram of a gated loop unit provided in an embodiment of this application;

[0053] Figure 8 is a schematic diagram of the training process of a prediction model provided in an embodiment of this application;

[0054] Figure 9 is a schematic diagram of a welding parameter prediction device provided in an embodiment of this application;

[0055] Figure 10 is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0056] The embodiments of this application are described below with reference to the accompanying drawings. It should be understood that the embodiments described below with reference to the accompanying drawings are exemplary descriptions for explaining the technical solutions of the embodiments of this application, and do not constitute a limitation on the technical solutions of the embodiments of this application.

[0057] Those skilled in the art will understand that, unless specifically stated otherwise, the singular forms “a,” “an,” “the,” and “the” used herein may also include the plural forms. It should be further understood that the terms “comprising” and “including” as used in embodiments of this application mean that the corresponding feature can be implemented as the presented feature, information, data, step, operation, element, and / or component, but do not exclude implementation as other features, information, data, step, operation, element, component, and / or combinations thereof supported by the art. It should be understood that when we say that an element is “connected” or “coupled” to another element, the one element can be directly connected or coupled to the other element, or it can mean that the one element and the other element establish a connection relationship through an intermediate element. Furthermore, “connected” or “coupled” as used herein can include wireless connection or wireless coupling. The term “and / or” as used herein indicates at least one of the items defined by the term; for example, “A and / or B” can be implemented as “A,” or as “B,” or as “A and B.”

[0058] In this application embodiment, the terms "module" or "unit" refer to a computer program or part of a computer program that has a predetermined function and works with other related parts to achieve a predetermined goal, and can be implemented wholly or partially using software, hardware (such as processing circuitry or memory), or a combination thereof. Similarly, a processor (or multiple processors or memory) can be used to implement one or more modules or units. Furthermore, each module or unit can be part of an overall module or unit that includes the functionality of that module or unit.

[0059] To make the objectives, technical solutions, and advantages of this application clearer, the embodiments of this application will be described in further detail below with reference to the accompanying drawings.

[0060] Ultrasonic welding of electrode tabs is a crucial step in battery cell production, and its welding quality directly affects the cell's performance, safety, and production costs. Welding time and welding power are two key factors in selecting welding parameters.

[0061] In related technologies, key data such as welding time and welding power can only be obtained after actual welding experiments. Because welding time and welding power can only be obtained afterward, the identification and resolution of welding problems are delayed. Therefore, predicting welding time and welding power and intervening in a timely manner based on the prediction results to improve battery welding quality has become a top priority.

[0062] The welding parameter prediction method, apparatus, electronic equipment, and storage medium provided in this application are intended to solve the above-mentioned technical problems in related technologies.

[0063] The technical solutions of this application and their effects are described below through several exemplary embodiments. It should be noted that the following embodiments can be referenced, borrowed from, or combined with each other. Identical terms, similar features, and similar implementation steps in different embodiments will not be repeated.

[0064] It is understood that in the welding parameter prediction method provided in the embodiments of this application, any step of the method can be executed by an electronic device and / or a server. All steps in the method can be executed independently by an electronic device or a server, or jointly by an electronic device and a server.

[0065] The server can be a standalone physical server, a server cluster or distributed system consisting of multiple physical servers, or a cloud server providing cloud computing services. Electronic devices can be smartphones, tablets, laptops, desktop computers, smart voice interaction devices (such as smart speakers), wearable electronic devices (such as smartwatches), in-vehicle terminals, smart home appliances (such as smart TVs), AR / VR devices, etc., but are not limited to these.

[0066] The embodiments of this application will subsequently be described using a server as the execution entity; however, this does not constitute a limitation on the embodiments of this application. The method provided in the embodiments of this application, in addition to the welding scenario during battery cell manufacturing, can also be applied to other scenarios for predicting at least one of the welding time and welding power during the welding process.

[0067] Figure 1 is a flowchart illustrating a welding parameter prediction method provided in an embodiment of this application. As shown in the figure, the method includes:

[0068] S101. Obtain the characteristic value of the first welding parameter at the first moment; wherein, the first welding parameter includes at least one of the material parameter of the welding operation object and the welding process parameter.

[0069] In this embodiment of the application, the object of the welding operation can be a battery cell. The first welding parameters and the second welding parameters provided in this embodiment of the application will be described in detail below, taking the battery cell as an example.

[0070] Specifically, in the embodiments of this application, the material parameters include, but are not limited to: the number of tab layers of the battery cell, the foil thickness, the solder depth, and the welding area; the welding process parameters include, but are not limited to: energy, pressure, and amplitude. It should be noted that, in the embodiments of this application, the specific parameters included in the material parameters and welding process parameters can be limited according to actual conditions, and the embodiments of this application do not limit the specific parameter types of the material parameters and welding process parameters.

[0071] Specifically, in this embodiment of the application, the welding process of the battery cell includes multiple consecutive moments. In this embodiment of the application, the characteristic value of the first welding parameter at any one moment is collected from the multiple consecutive moments, and the collected moment is taken as the first moment.

[0072] Optionally, in the embodiments of this application, the characteristic value of the first welding parameter at the first moment can be obtained by means of automatic recording by the welding device, or by measuring and recording the characteristic value of the first welding parameter by a set monitoring device. The specific acquisition method can be determined according to actual needs.

[0073] S102. Input the feature values ​​of the first welding parameter into the pre-trained prediction model;

[0074] The prediction model is trained using the parameter values ​​of the first welding parameter at multiple sample times as training samples, and is trained through a graph convolutional network and a gated recurrent network. Each training sample also includes a training label, which is the feature value of the second welding parameter at the next time step of the corresponding sample time. The second welding parameter includes at least one of welding time and welding power, and the first welding parameter is the independent variable of the second welding parameter.

[0075] In this embodiment, the first welding parameter is the independent variable of the second welding parameter; in other words, there is a causal relationship between the first and second welding parameters. For example, the number of tab layers and the foil thickness affect the heat conduction and energy required during welding. A larger number of tab layers and a thicker foil require a larger welding area or higher power to ensure a strong connection. An increased welding area may require a longer welding time or a higher amplitude to cover a larger area. In ultrasonic welding, amplitude and pressure directly affect the welding energy, which is related to welding time and power. A higher amplitude may require higher power, while greater pressure may shorten the welding time. Therefore, given the causal relationship between the first and second welding parameters, this embodiment inputs the feature value of the first welding parameter at a first moment into a pre-trained model to obtain the feature value of the second welding parameter.

[0076] Specifically, the prediction model is obtained through pre-training. In this embodiment, the prediction model includes a graph convolutional network and a gated recurrent network. The graph convolutional network extracts spatial features of the input data through graph convolution operations, effectively modeling the relationships between the first welding parameters, such as the interaction between the number of tab layers and the foil thickness. The gated recurrent network processes parameter changes through a gating mechanism, capturing the dynamic changes of parameters during the welding process. By combining the spatial feature extraction capability of the graph convolutional network and the dynamic information processing capability of the gated recurrent network, this embodiment can more comprehensively capture the features of the input data, thereby significantly improving the prediction performance of welding time and welding power.

[0077] Specifically, in this embodiment of the application, during the pre-training of the prediction model, feature values ​​of the first welding parameter at multiple moments during the welding process are collected in advance as sample data; simultaneously, feature values ​​of the second welding parameter at the next moment at multiple consecutive moments are collected as sample labels. The prediction model is trained based on the sample data and sample labels to obtain a trained prediction model. This trained prediction model can output the feature values ​​of the second welding parameter at the next moment based on the input feature values ​​of the first welding parameter at the first moment.

[0078] S103. Obtain the eigenvalues ​​of the second welding parameters at the next time step after the first time step output by the prediction model.

[0079] In this embodiment of the application, the feature value of the first welding parameter at the first moment is input into the prediction model to obtain the feature value of the second welding parameter at the next moment after the first moment. For example, if the first moment is 10:20:00 and the time step is 1 second (i.e., welding once per second), then the next moment after the first moment is 10:20:01. That is, the prediction model outputs the feature value of the second welding parameter at 10:20:01.

[0080] Optionally, since cell welding is a continuous process, training the prediction model using sample data and labels from a single moment may result in errors and fail to reflect the changing trends of welding time and welding power. Therefore, in this embodiment, feature values ​​of the first welding parameters at multiple consecutive moments are collected, and the feature values ​​of the first welding parameters at each moment are sequentially input into the prediction model according to the chronological order to obtain the feature values ​​of the second welding parameters at the next moment. This better reflects the changing trends of the feature values ​​of the second welding parameters, allowing workers to intervene in the welding process based on the changing trends of the second welding parameters and improve welding quality.

[0081] Optionally, in this embodiment of the application, after obtaining the characteristic value of the second welding parameter at the next moment of the first moment, the characteristic value of the second welding parameter at the next moment can be compared with the preset range of the second welding parameter. If the characteristic value of the second welding parameter at the next moment is within the preset range, no intervention is required. If the characteristic value of the second welding parameter at the next moment is not within the preset range, at least one of the material parameters and welding process parameters needs to be adjusted.

[0082] In this embodiment, the prediction model is trained using the parameter values ​​of the first welding parameter at multiple sample times as training samples, and is trained through a graph convolutional network and a gated recurrent network. The graph convolutional network is good at capturing the spatial features of the data, while the gated recurrent network is good at capturing the temporal features of the data. Therefore, the prediction model can more accurately reflect the complex relationship between welding parameters, which makes the feature value of the output second welding parameter more accurate. This helps to achieve accurate prediction of the welding process and solves the technical problem in related technologies where welding time and welding power can only be obtained after the fact, which leads to the delay in the identification and resolution of welding problems.

[0083] Based on the above embodiments, as an optional embodiment, the graph convolutional network includes: a graph convolution module and an attention module; the gated recurrent network includes: a feature extraction module and a second welding parameter prediction module;

[0084] The feature values ​​of the first welding parameter are input into a pre-trained prediction model to obtain the feature values ​​of the second welding parameter at the next time step after the first time step output by the prediction model, including:

[0085] The feature values ​​of the first welding parameter are input into the graph convolution module and the attention module respectively to obtain the target feature matrix output by the graph convolution module and the target feature vector output by the attention module; based on the target feature matrix and the target feature vector, the updated feature values ​​of the first welding parameter are obtained.

[0086] The updated feature values ​​of the first welding parameters are input into the feature extraction module to obtain the target hidden state output by the feature extraction module. The target hidden state indicates the dependency relationship between the first welding parameters. The target hidden state is input into the second welding parameter prediction module to output the feature values ​​of the second welding parameters at the next time step.

[0087] Specifically, Figure 2 is a schematic diagram of the structure of a prediction model provided in an embodiment of this application. As shown in the figure, the prediction model 20 includes a graph convolutional network 201 and a gated recurrent network 202 in sequence. The graph convolutional network 201 includes a graph convolution module 30 and an attention module 40, and the gated recurrent network 202 includes a feature extraction module 50 and a second welding parameter prediction module 60. The feature value of the first welding parameter at the first time is input into the prediction model 20. After processing by the graph convolutional network 201 and the gated recurrent network 202, the feature value of the second welding parameter at the next time after the first time is output.

[0088] In this embodiment, the feature values ​​of the first welding parameters are input to the graph convolution module 30 and the attention module 40, respectively. The graph convolution module 30 uses the interrelationships between the first welding parameters to extract and fuse features to form a target feature matrix, which reflects the interrelationships of the welding parameters in the graph structure. The attention module 40 outputs a target feature vector based on the weights of different feature channels. This target feature vector highlights the feature channels most critical for predicting the welding parameters at the next moment. Then, the target feature matrix output by the graph convolution module 30 and the target feature vector output by the attention module 40 are subjected to a dot product operation. The result of the dot product operation is used to update the feature values ​​of the first welding parameters. The updated feature values ​​of the first welding parameters are input to the feature extraction module 50, which further processes these features and outputs the target hidden state. The target hidden state not only contains the feature information of the welding parameters themselves, but also implies the dependencies between these parameters, which is crucial for understanding the complex interactions in the welding process. Finally, the target hidden state is input into the second welding parameter prediction module 60, which uses the information in the hidden state to predict the feature value of the second welding parameter at the next moment. Since the hidden state has captured the dependencies and key features between the welding parameters, the prediction result can more accurately reflect the changing trend of the second welding parameter.

[0089] This application embodiment effectively extracts and utilizes the feature information of welding parameters by combining graph convolution modules and attention modules, thereby improving the prediction accuracy of welding parameters at the next moment.

[0090] Based on the above embodiments, as an optional embodiment, the graph convolution module includes at least one graph convolution unit; there are predetermined interrelationships among the first welding parameters;

[0091] The feature values ​​of the first welding parameter are input into the graph convolution module to obtain the target feature matrix output by the graph convolution module, including:

[0092] Based on the characteristic values ​​of the first welding parameters, a feature map is determined at the first moment; wherein, the nodes of the feature map are the first welding parameters, and the edges between the nodes are the pre-determined relationships between the first welding parameters;

[0093] The feature maps are sequentially input into each graph convolutional unit to obtain the feature matrices output by each graph convolutional unit, and the feature matrix output by the last graph convolutional unit is used as the target feature matrix.

[0094] Specifically, Figure 3 is a schematic diagram of the structure of a graph convolution module provided in an embodiment of this application. As shown in Figure 3, the graph convolution module 30 includes at least one graph convolution unit 301. Each graph convolution unit 301 is connected in series. The feature value of the first welding parameter at the first moment is input into the graph convolution module 30. After passing through the convolution operation of each graph convolution unit 301 in sequence, the last graph convolution unit 301 outputs the target feature matrix.

[0095] Specifically, in this embodiment, the first welding parameter is used as a node of the feature map, and edges between nodes are constructed according to the predetermined interrelationships between the welding parameters. In this way, the feature map intuitively represents the welding parameters and their interrelationships. The constructed feature map is sequentially input into each graph convolution unit 301. The graph convolution unit 301 can extract feature information from the feature map using convolution operations. Through the stacking of multiple graph convolution units 301, more complex features can be iteratively extracted from low-level features. Each graph convolution unit 301 outputs a feature matrix, which contains the feature information extracted from the feature map. The feature matrix output by the last graph convolution unit 301 is used as the target feature matrix. The target feature matrix integrates the processing results of all graph convolution units 301 and contains a high-level feature representation of the welding parameters.

[0096] It should be noted that in this embodiment, after obtaining the feature values ​​of the first welding parameters at the first moment, it is necessary to generate a corresponding feature map based on the feature values ​​of the first welding parameters. The nodes in the feature map are the first welding parameters, and the edges between the nodes are the relationships between the first welding parameters. For example, the first welding parameters include: the number of tab layers of the battery cell, the foil thickness, the weld depth, the welding area, the energy, the pressure, and the amplitude. These welding parameters are used as nodes in the feature map, and the relationships between these welding parameters are used as edges between the corresponding nodes. It should be noted that in this embodiment, the relationships between the first welding parameters are predetermined. For example, the relationship between the number of tab layers and the foil thickness means that there is an edge between the node number of tab layers and the node foil thickness.

[0097] Furthermore, based on the feature map, the feature matrix X, adjacency matrix A, and degree matrix D of each node can be obtained; where the feature matrix A is used to describe the initial feature values ​​of each node, the adjacency matrix X is used to describe the connection relationship between each node, and the degree matrix is ​​used to describe the number of edges connected to each node.

[0098] Furthermore, the standardized adjacency features of the feature map can be obtained. The formula (1) for the standardized adjacency matrix is ​​shown below:

[0099] in, yes The degree matrix, It is obtained by adding the self-loop matrix I to the adjacency matrix A, as shown in formula (2) below:

[0100] Furthermore, the feature map is input into the first graph convolutional unit 301, and the output feature matrix H is... (1) , where the characteristic matrix H (1) The calculation formula (3) is shown below:

[0101] Among them, W (0) σ is the weight matrix of the first graph convolutional unit 301, and σ is the activation function (e.g., ReLU).

[0102] When the number of graph convolutional units 301 is greater than 1, the feature matrix H output by the first graph convolutional unit 301 is... (1) The input is fed into each subsequent graph convolutional unit 301, and the feature matrix output by each graph convolutional unit 301 is obtained in sequence. In this embodiment, the feature matrix output by the last graph convolutional unit 301 is used as the target feature matrix; wherein, the calculation formula (4) of the target feature matrix is ​​as follows:

[0103] Among them, H (l+1) H is the feature matrix output by the (l+1)th graph convolutional unit 301; (l) W is the feature matrix output by the l-th graph convolutional unit 301. (l) Let be the weight matrix of the l-th graph convolutional unit 301. When l equals 1, let be the feature matrix output by the second graph convolutional unit 301; when l equals 2, let be the feature matrix output by the third graph convolutional unit 301, and so on, until the feature matrix output by the last graph convolutional unit 301 is obtained, which is used as the target feature matrix.

[0104] In this embodiment, the feature map is processed by graph convolution units, which can effectively extract feature information from the first welding parameters. At the same time, the target feature matrix contains high-level feature representations of the welding parameters, which is of great significance for subsequent tasks such as prediction of the second welding parameters and optimization of the welding process.

[0105] Based on the above embodiments, as an optional embodiment, the attention module includes at least one attention unit and a global pooling unit; wherein, each attention unit includes an attention channel corresponding to the first welding parameter;

[0106] The feature values ​​of the first welding parameter are input into the attention module to obtain the target feature vector output by the attention module, including:

[0107] Determine the input data for each attention unit; for the first attention unit, the input data is the feature value of the first welding parameter; for non-first attention units, the input data is the feature value output after convolution calculation of the previous attention unit.

[0108] For each attention unit, the input data corresponding to each first welding parameter is input into the corresponding attention channel, and convolution calculation is performed in the corresponding attention channel to obtain the feature value output by each attention channel; wherein, the weight value of each attention channel is different;

[0109] The feature values ​​output by each attention channel in the last attention unit are input into the global pooling unit. The maximum value among all feature values ​​in each attention channel is extracted as the corresponding feature value after updating the first welding parameters. The feature values ​​after updating the first welding parameters are combined into a one-dimensional feature vector, which is used as the target feature vector.

[0110] Specifically, Figure 4 is a schematic diagram of the structure of an attention module provided in an embodiment of this application. As shown in Figure 4, the attention module 40 includes at least one attention unit 401 and a global pooling unit 402; wherein the attention unit 401 is connected in sequence and then connected in series with the global pooling unit 402. The feature value of the first welding parameter at the first moment is input into the attention module 40, and the target feature vector is output.

[0111] Specifically, in this embodiment, for the first attention unit 401, the input data is the feature value of the first welding parameter; for subsequent attention units 401, the input data is the feature value output by the convolution calculation of the previous attention unit 401. This design allows each attention unit 401 to perform further feature extraction and updating based on the processing result of the previous unit. Furthermore, each attention unit 401 contains multiple attention channels, each corresponding to a different weight value. The input data corresponding to each first welding parameter is input into the corresponding attention channel, and convolution calculation is performed in each attention channel to obtain the feature value output by that channel. Since the weight values ​​of each channel are different, the features extracted by each channel will also be different. Then, the feature values ​​output by each attention channel in the last attention unit 401 are input into the global pooling unit 402. The global pooling unit 402 extracts the maximum value among all feature values ​​in each attention channel as the corresponding feature value after updating the first welding parameters. This design helps to retain the most significant feature information in each channel. Finally, the feature values ​​after updating the first welding parameters are combined into a one-dimensional feature vector. This feature vector contains the updated feature values ​​of all welding parameters after being processed by multiple attention units 401 and global pooling unit 402, and can be regarded as the target feature vector.

[0112] In this embodiment, through the processing of multiple attention units and attention channels, the feature values ​​of the first welding parameters can be extracted and updated step by step, making the final feature vector more accurate and comprehensive. At the same time, the different weight values ​​of each attention channel help the model focus on different feature information, improving the model's generalization ability and robustness. In addition, the global pooling unit extracts the maximum value in each channel, which helps to retain the most significant feature information while reducing the dimension of the feature vector and lowering the computational complexity. The final target feature vector contains all the updated feature values ​​of the first welding parameters, effectively improving the prediction accuracy of the subsequent second welding parameters.

[0113] Based on the above embodiments, as an optional embodiment, the feature extraction module includes multiple gated loop units and a fully connected unit;

[0114] The updated feature values ​​of the first welding parameters are input into the feature extraction module to obtain the target hidden state output by the feature extraction module, including:

[0115] For the first gated loop unit, the input data is the feature value after the first welding parameter is updated, and the output data is the hidden state of the first gated loop unit;

[0116] For non-first gated loop units, the input data are the feature values ​​after the first welding parameter update and the hidden state output by the previous gated loop unit, and the output is the hidden state of the corresponding gated loop unit.

[0117] The hidden state output by the last gated loop unit is input to the fully connected unit for full connection processing, and the hidden state output by the fully connected unit is obtained as the target hidden state.

[0118] Specifically, Figure 5 is a schematic diagram of the structure of a feature extraction module provided in an embodiment of this application. As shown in Figure 5, the feature extraction module 50 includes multiple gated loop units 501 and a fully connected unit 502. The input data of each gated loop unit 501 is the parameter value of the first welding parameter after the first time step and the hidden state output by the previous gated loop unit 501. The last gated loop unit 501 and the fully connected unit 502 are connected in series. The last gated loop unit inputs the output hidden state to the fully connected unit 502, and the fully connected unit 502 outputs the target hidden state.

[0119] Specifically, for the first gated loop unit 501, its input is the updated feature value of the first welding parameters. The first welding parameters include at least one of material parameters and welding process parameters, and their feature values ​​are obtained after preprocessing and feature extraction. For the first gated loop unit 501, the output is the hidden state of the first gated loop unit 501, which contains information from the feature values ​​of the first welding parameters.

[0120] Specifically, for a non-first gated loop unit 501, its input includes two parts: first, the updated feature value of the first welding parameters (this part is the same in all gated loop units 501); and second, the hidden state output by the previous gated loop unit 501. This design allows each gated loop unit 501 to utilize the information from the previous unit, thereby transmitting and integrating information throughout the sequence. Furthermore, the output of a non-first gated loop unit 501 is the hidden state of the current gated loop unit 501, which contains all relevant information from the beginning of the sequence to the current position.

[0121] Specifically, the hidden state output by the last gated recurrent unit 501 is input to the fully connected unit 502. The fully connected unit 502 further processes the hidden state, typically by using linear transformations and activation functions (such as ReLU, sigmoid, etc.) to extract higher-level features. The output of the fully connected unit 502 is used as the target hidden state, which can be used for subsequent tasks such as classification, regression, or sequence generation.

[0122] In this embodiment, each gated loop unit outputs a hidden state, which is a deep representation of the input data. As the data is passed between the gated loop units, the features are gradually extracted and deepened. The target hidden state output by the fully connected unit can be used as the input for subsequent tasks. Since the fully connected layer has high flexibility, it can adapt to different task requirements.

[0123] Based on the above embodiments, as an optional embodiment, the second welding parameter prediction module includes multiple gated loop units and a global pooling unit;

[0124] The target hidden state is input into the second welding parameter prediction module, which outputs the feature values ​​of the second welding parameters at the next time step, including:

[0125] Determine the input data for each gated loop unit; for the first gated loop unit, the input data is the target's hidden state; for subsequent gated loop units, the input data is the hidden state output by the previous gated loop unit.

[0126] For each gated loop unit, perform a dot product operation on the hidden state of its own output and output the predicted value of the second welding parameter;

[0127] The predicted values ​​of the second welding parameters output by each gated loop unit are input into the global pooling unit. The predicted values ​​of all the second welding parameters are pooled to obtain the feature values ​​of the second welding parameters at the next time step.

[0128] Specifically, Figure 6 is a schematic diagram of the structure of a second welding parameter prediction module provided in an embodiment of this application. As shown in Figure 6, the second welding parameter prediction module 60 includes multiple gated loop units 601 and a global pooling unit 602. Each gated loop unit 601 inputs the result of the dot product operation of its own output hidden state into the global pooling unit 602. At the same time, the hidden state output by the preceding gated loop unit 601 is used as the input hidden state of the subsequent gated loop unit 601.

[0129] Specifically, in this embodiment, for the first gated loop unit 601, the input data is the target hidden state, which may contain prior information or context information related to the welding task; for non-first gated loop units 601, the input data is the hidden state output by the previous gated loop unit 601. In this way, each gated loop unit 601 can use the information of the previous unit to update its hidden state.

[0130] Specifically, in this embodiment, each gated loop unit 601 outputs an updated hidden state and performs a dot product operation on the output hidden state to generate a predicted value for the second welding parameter from the hidden state. The predicted values ​​of the second welding parameter output by each gated loop unit 601 are input into the global pooling unit 602, and all predicted values ​​are summarized or averaged to generate a single, representative predicted value, which serves as the feature value of the second welding parameter at the next time step.

[0131] In this embodiment, the gated loop unit can integrate information in the sequence and capture the trend of welding parameters changing over time. The global pooling unit summarizes all predicted values, which helps to generate the feature value of the second welding parameter at the next moment.

[0132] Figure 7 is a schematic diagram of a gated loop unit provided in an embodiment of this application. As shown in the figure, the gated loop unit includes a reset gate and an update gate. For the current gated loop unit, based on the input data x... t The hidden state h output by the previous gated loop unit t-1 The gate r is reset by calculating the current gate control cycle unit. t and update gate z t Reset door r t Formula (5) and the z of the update gate t Formula (6) is shown below: r t =σ(W r ·[h t-1 ,x t ]+b r (5) z t =σ(W z ·[h t-1 ,x t ]+b z (6)

[0133] In the formula, σ represents the fully connected layer and the sigmoid activation function, W r and W z The weights for updating the gate and resetting the gate are respectively, b. r and b z These are the offsets for resetting the door and updating the door, respectively.

[0134] According to the reset gate r t Target input x t The hidden state h output by the previous gated loop unit t-1 Get the candidate hidden state h t ′, reset door r t The hidden state h of the output of the previous gating loop unit is controlled. t-1 How to flow into the candidate hidden state h of the current gated loop unit t ′, while the hidden state h output by the previous gated loop unit t-1 It contains all historical information about the characteristic values ​​of the first welding parameter, therefore the gate r is reset. t It can be used to discard historical information that is irrelevant to the prediction.

[0135] According to the updated gate z t The hidden state h output by the previous gated loop unit t-1 and candidate hidden state h t ′, obtain the hidden state h output by the current gated loop unit. t Update Gate Z t It can control how the hidden state should be selected as a candidate hidden state h t Updated by ′.

[0136] Formula (7) for the candidate hidden state and the hidden state h output by the current gated loop unit t Formula (8) is shown below: h′ t =tanh(W h ·[r t *h t-1 ,x t ]) (7) h t =(1-z) t )*h t-1 +z t *h′ t (8)

[0137] In the formula, tanh refers to the hyperbolic tangent function of the activation function, W h The weights corresponding to the candidate hidden states are used when calculating them.

[0138] Specifically, the input data of each gated loop unit in the feature extraction module is the feature value of the first welding parameter at the first moment, that is: the x of each gated loop unit in the above formulas (5) to (7). t All are characteristic values ​​of the first welding parameters at the first moment; further, in the embodiments of this application, starting from the second gated loop unit, h t-1This represents the hidden state output by the previous gated loop unit. Here, "t" can be understood as the sequence number of the gated loop unit.

[0139] Specifically, in the second welding parameter prediction module, the input of each gated loop unit is only a hidden state. In other words, in the embodiments of this application, x in the above formulas (5) to (7) is... t This can be considered as 0, which means that the embodiments of this application have improved the conventional gated loop unit.

[0140] It should be noted that for the first gated loop unit, since there is no previous gated loop unit, the hidden state of the output of the previous gated loop unit can be initialized, that is, the initial hidden state is determined. Usually, the initial hidden state can be set to a zero vector or a random small value vector.

[0141] Based on the above embodiments, as an optional embodiment, the prediction model is generated in the following manner, including:

[0142] The parameter values ​​of the first welding parameter and the second welding parameter at multiple time points are obtained in advance. The parameter values ​​of the first welding parameter and the second welding parameter at each time point are preprocessed to obtain the feature values ​​of the first welding parameter and the second welding parameter at each time point. The preprocessing includes at least one of data cleaning and normalization.

[0143] Based on the characteristic values ​​of the first welding parameter and the second welding parameter at each time point, determine the characteristic values ​​of the first welding parameter at multiple sample times and the characteristic value of the second welding parameter at the next time point of the corresponding sample time.

[0144] The initial prediction model is trained by using the first welding parameter feature value at the sample time as the training sample and the second welding parameter feature value at the next time of the corresponding sample time as the training label until the training stop condition is met, thus obtaining the welding prediction model.

[0145] Specifically, Figure 8 is a schematic diagram of the training process of a prediction model provided in an embodiment of this application. As shown in the figure, in this embodiment of the application, the parameter values ​​of the first welding parameters (such as material parameters and welding process parameters) at multiple times and the parameter values ​​of the second welding parameters (such as welding time and welding power) at the next time at multiple times are collected in advance as sample data. These parameter values ​​should cover data under different welding conditions to ensure the generalization ability of the model.

[0146] Specifically, the collected sample data is preprocessed, including cleaning to remove outliers and missing values ​​to ensure the accuracy and completeness of the data. Optionally, the preprocessing of the sample data also includes normalization to convert parameter values ​​of different dimensions to the same scale to improve the training efficiency and prediction accuracy of the model. The normalization method can be selected according to the characteristics of the data, such as min-max normalization, Z-score standardization, etc.

[0147] Specifically, based on the preprocessed data, the feature values ​​of the first welding parameter at each time point and the feature values ​​of the second welding parameter at the next time point are determined. The feature values ​​of the first welding parameter at the sample time point are used as training samples, and the feature values ​​of the second welding parameter at the next time point are used as training labels. The sample set constructed in this way can reflect the relationship between welding parameters and time, providing a basis for predicting the welding result at the next time point.

[0148] Specifically, in this embodiment, graph convolutional networks and gated recurrent networks are used as the initial prediction models. The constructed sample set is input into the prediction model to be trained, and the model parameters are continuously adjusted to reduce the prediction error. When the prediction model meets the training stopping conditions (such as the prediction error reaching a preset threshold, the number of training iterations reaching an upper limit, etc.), training stops and the final prediction model is obtained.

[0149] Specifically, after obtaining the final prediction model, the feature value of the first parameter at the first moment is input into the prediction model to obtain the feature value of the second welding parameter at the next moment after the first moment.

[0150] Optionally, the Adam algorithm can be used to train the prediction model during the training phase. The Adam algorithm is one of the commonly used optimization algorithms in deep neural networks. It combines the advantages of momentum optimization and adaptive learning rate, effectively updating network parameters and accelerating model convergence. The core of Adam is to combine the first moment (mean) and second moment (variance) estimates of the gradient, calculate these two using an exponentially weighted moving average, and correct the bias of the estimates.

[0151] Optionally, this application embodiment uses mean squared error (MSE) as a loss function to measure the difference between the model's predicted value and the actual value. Mean squared error is commonly used in regression problems and is suitable for situations requiring the prediction of continuous values. It calculates the average of the squares of the differences between the predicted and actual values, thereby quantifying the magnitude of the prediction error and helping to optimize the model to minimize prediction error and improve prediction accuracy. The formula (9) for calculating mean squared error (MSE) is shown below:

[0152] In the formula, MSE represents the mean squared error loss function, y i Indicates the actual value. This represents the predicted value, where n is the number of samples.

[0153] Optionally, embodiments of this application also employ R 2 As a convergence criterion for the model, the mean squared error (MSE) measures the difference between the model's predicted and actual values, while R0 is used for the model's convergence. 2 Evaluate the model's ability to explain the variance of the data. In regression tasks, R... 2 R is used to measure the goodness of fit of a model, and its value ranges from 0 to 1. The closer the value is to 1, the better the model fit. By introducing R... 2 As a convergence criterion, it can monitor the fitting of the model during training and evaluate the model's predictive ability, thereby further optimizing the model and improving its prediction accuracy and stability in practical applications. 2 The calculation formula (10) is shown below:

[0154] Among them, SS res This is denoted as the residual sum of squares, used to describe the sum of squares of the differences between the model's predicted and actual values; SS tot It is denoted as the total sum of squares of deviations and is used to describe the sum of squares of the differences between the actual value and the mean.

[0155] Specifically, in this embodiment of the application, based on the prediction model obtained through training, the feature value of the first welding parameter at the current moment can be used as the input of the model to obtain the predicted value of the second welding parameter at the next moment output by the model, and based on the predicted value of the second welding parameter output, it can be determined whether the welding process needs to be intervened to improve the welding quality.

[0156] In this embodiment, data preprocessing and feature extraction reduce the impact of noise and redundant information on the model, thereby improving the accuracy of prediction. At the same time, the prediction model can help identify and solve potential problems in advance, improve welding efficiency and quality, reduce scrap rate and rework rate, and thus reduce production costs.

[0157] Figure 9 is a schematic diagram of a welding parameter prediction device provided in an embodiment of this application. As shown in the figure, the device may include: a first processing module 901, a second processing module 902 and a third processing module 903.

[0158] The first processing module 901 is configured to acquire the feature values ​​of the first welding parameters at a first moment; wherein the first welding parameters include at least one of the material parameters of the welding operation object and the welding process parameters.

[0159] The second processing module 902 is configured to input the feature values ​​of the first welding parameter into a pre-trained prediction model; wherein the prediction model is trained using the parameter values ​​of the first welding parameter at multiple sample times as training samples, through a graph convolutional network and a gated recurrent network; each training sample also includes a training label, which is the feature value of the second welding parameter at the next time step of the corresponding sample time; wherein the second welding parameter includes at least one of welding time and welding power, and the first welding parameter is the independent variable of the second welding parameter;

[0160] The third processing module 903 is configured to obtain the feature values ​​of the second welding parameters at the next time step after the first time step output by the prediction model.

[0161] The welding parameter prediction device of this application embodiment can execute the welding parameter prediction method provided in this application embodiment. The implementation principle is similar. The actions performed by each module in the welding parameter prediction device of each embodiment of this application are corresponding to the steps in the welding parameter prediction method of each embodiment of this application. For detailed functional descriptions of each module of the welding parameter prediction device, please refer to the descriptions in the corresponding methods shown above, which will not be repeated here.

[0162] In this embodiment, the parameter values ​​of the first welding parameter at multiple sample times are used as training samples, and the feature values ​​of the second welding parameter at the next time step of the corresponding sample time are used as training labels to train the graph convolutional network and gated recurrent units to obtain the prediction model. The feature values ​​of the first welding parameter at the first time step are obtained and input into the pre-trained prediction model to obtain the feature values ​​of the second welding parameter at the next time step of the first time step output by the prediction model. Compared to the traditional method of obtaining welding time and welding power through actual welding experiments, this embodiment can accurately predict the welding time and welding power of the battery cell, effectively improving the accuracy of the welding time and welding power prediction results, and instructing personnel to intervene in a timely manner based on the prediction results to improve welding quality.

[0163] Figure 10 is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. As shown in Figure 10, the electronic device 4000 includes a processor 4001 and a memory 4003. The processor 4001 and the memory 4003 are connected, for example, via a bus 4002. Optionally, the electronic device 4000 may further include a transceiver 4004, which can be used for data interaction between the electronic device and other electronic devices, such as sending and / or receiving data. It should be noted that in practical applications, the transceiver 4004 is not limited to one unit, and the structure of the electronic device 4000 does not constitute a limitation on the embodiments of this application.

[0164] Processor 4001 may be a CPU (Central Processing Unit), a general-purpose processor, a DSP (Digital Signal Processor), an ASIC (Application Specific Integrated Circuit), an FPGA (Field Programmable Gate Array), or other programmable logic devices, transistor logic devices, hardware components, or any combination thereof. It can implement or execute the various exemplary logic blocks, modules, and circuits described in conjunction with the disclosure of this application. Processor 4001 may also be a combination that implements computational functions, such as including one or more microprocessor combinations, a combination of a DSP and a microprocessor, etc.

[0165] Bus 4002 may include a pathway for transmitting information between the aforementioned components. Bus 4002 may be a PCI (Peripheral Component Interconnect) bus or an EISA (Extended Industry Standard Architecture) bus, etc. Bus 4002 can be divided into address bus, data bus, control bus, etc. For ease of illustration, only one thick line is used to represent it in Figure 10, but this does not indicate that there is only one bus or one type of bus.

[0166] The memory 4003 may be ROM (Read Only Memory) or other types of static storage devices capable of storing static information and instructions, RAM (Random Access Memory) or other types of dynamic storage devices capable of storing information and instructions, or EEPROM (Electrically Erasable Programmable Read Only Memory), CD-ROM (Compact Disc Read Only Memory) or other optical disc storage, optical disc storage (including compressed optical discs, laser discs, optical discs, digital universal optical discs, Blu-ray discs, etc.), magnetic disk storage media, other magnetic storage devices, or any other medium capable of carrying or storing computer programs and capable of being read by a computer, without limitation herein.

[0167] The memory 4003 stores computer programs that execute embodiments of this application, and its execution is controlled by the processor 4001. The processor 4001 executes the computer programs stored in the memory 4003 to implement the steps shown in the foregoing method embodiments.

[0168] The electronic device package may include, but is not limited to, mobile terminals such as mobile phones, laptops, digital broadcast receivers, PDAs (personal digital assistants), PADs (tablet computers), PMPs (portable multimedia players), and in-vehicle terminals (such as in-vehicle navigation terminals), as well as fixed terminals such as digital TVs and desktop computers. The electronic device shown in Figure 10 is merely an example and should not be construed as limiting the functionality and scope of use of the embodiments of this application.

[0169] This application provides a computer-readable storage medium storing a computer program. When executed by a processor, the computer program can implement the steps and corresponding content of the aforementioned method embodiments. Compared with related technologies, it can achieve the following:

[0170] It should be noted that the computer-readable medium described above in this application can be a computer-readable signal medium, a computer-readable medium, or any combination of the two. A computer-readable storage medium can be, for example,—but not limited to—an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of a computer-readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In this application, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. In this application, a computer-readable signal medium can include a data signal propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals can take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. A computer-readable signal medium can be any computer-readable medium other than a computer-readable storage medium, which can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device. The program code contained on the computer-readable medium can be transmitted using any suitable medium, including but not limited to: wires, optical fibers, RF (radio frequency), etc., or any suitable combination thereof.

[0171] This application also provides a computer program product, including a computer program that, when executed by a processor, can implement the steps and corresponding content of the aforementioned method embodiments. Compared with related technologies, it can achieve:

[0172] The terms "first," "second," "third," "fourth," "1," "2," etc. (if present) in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in a sequence other than that shown in the figures or text.

[0173] It should be understood that although arrows indicate various operation steps in the flowcharts of this application's embodiments, the order in which these steps are implemented is not limited to the order indicated by the arrows. Unless explicitly stated herein, in some implementation scenarios of this application's embodiments, the implementation steps in each flowchart can be executed in other orders as required. Furthermore, some or all steps in each flowchart, based on the actual implementation scenario, may include multiple sub-steps or multiple stages. Some or all of these sub-steps or stages can be executed at the same time, and each sub-step or stage can also be executed at different times. In scenarios where execution times differ, the execution order of these sub-steps or stages can be flexibly configured according to requirements, and this application's embodiments do not limit this.

[0174] The above are only optional implementation methods for some implementation scenarios of this application. It should be noted that for those skilled in the art, other similar implementation methods based on the technical concept of this application, without departing from the technical concept of this application, also fall within the protection scope of the embodiments of this application. Industrial applicability:

[0175] The solution provided in this application can be applied to the field of welding technology. In this application embodiment, by obtaining the feature value of the first welding parameter at a first moment; inputting the feature value of the first welding parameter into a pre-trained prediction model, the feature value of the second welding parameter at the next moment after the first moment is obtained from the prediction model output; wherein, the second welding parameter includes at least one of welding time and welding power, and the first welding parameter is the independent variable of the second welding parameter, the accurate prediction of the welding time and welding power of the battery cell is realized, effectively improving the accuracy of the prediction results of welding time and welding power. At the same time, the welding process can be intervened in a timely manner according to the prediction results to improve the welding quality.

Claims

1. A method for predicting welding parameters, the method comprising: Obtain the feature value of the first welding parameter at the first moment; wherein, the first welding parameter includes at least one of the material parameter of the welding operation object and the welding process parameter; The feature value of the first welding parameter is input into a pre-trained prediction model; wherein the prediction model is trained using the parameter values ​​of the first welding parameter at multiple sample times as training samples, through a graph convolutional network and a gated recurrent network; each training sample also includes a training label, the training label being the feature value of the second welding parameter at the next time of the corresponding sample time; wherein the second welding parameter includes at least one of welding time and welding power, and the first welding parameter is the independent variable of the second welding parameter; The feature values ​​of the second welding parameters at the next time step after the first time step are obtained from the output of the prediction model.

2. The welding parameter prediction method according to claim 1, wherein, The graph convolutional network includes a graph convolution module and an attention module; the gated recurrent network includes a feature extraction module and a second welding parameter prediction module. The feature values ​​of the first welding parameter are input into a pre-trained prediction model to obtain the feature values ​​of the second welding parameter at the next time step after the first time step, as output by the prediction model, including: The feature values ​​of the first welding parameter are input into the graph convolution module and the attention module respectively to obtain the target feature matrix output by the graph convolution module and the target feature vector output by the attention module; the updated feature values ​​of the first welding parameter are obtained based on the target feature matrix and the target feature vector. The updated feature values ​​of the first welding parameters are input into the feature extraction module to obtain the target hidden state output by the feature extraction module. The target hidden state indicates the dependency relationship between the first welding parameters. The target hidden state is input into the second welding parameter prediction module to output the feature values ​​of the second welding parameters at the next moment.

3. The welding parameter prediction method according to claim 2, wherein, The graph convolution module includes at least one graph convolution unit; there are predetermined interrelationships among the first welding parameters; The step of inputting the feature values ​​of the first welding parameter into the graph convolution module to obtain the target feature matrix output by the graph convolution module includes: Based on the characteristic values ​​of the first welding parameters, a feature map is determined at the first moment; wherein, the nodes of the feature map are the first welding parameters, and the edges between the nodes are the predetermined relationships between the first welding parameters. The feature maps are sequentially input into each graph convolutional unit to obtain the feature matrices output by each graph convolutional unit, and the feature matrix output by the last graph convolutional unit is used as the target feature matrix.

4. The welding parameter prediction method according to claim 3, wherein, The attention module includes at least one attention unit and a global pooling unit; wherein each attention unit includes an attention channel corresponding to the first welding parameter; The step of inputting the feature values ​​of the first welding parameter into the attention module to obtain the target feature vector output by the attention module includes: Determine the input data for each attention unit; where, for the first attention unit, the input data is the feature value of the first welding parameter; for non-first attention units, the input data is the feature value output after convolution calculation of the previous attention unit; For each attention unit, the input data corresponding to each first welding parameter is input into the corresponding attention channel, and convolution calculation is performed in the corresponding attention channel to obtain the feature value output by each attention channel; wherein, the weight value of each attention channel is different; The feature values ​​output by each attention channel in the last attention unit are input into the global pooling unit. The maximum value among all feature values ​​in each attention channel is extracted as the corresponding feature value after updating the first welding parameters. The feature values ​​after updating the first welding parameters are combined into a one-dimensional feature vector, which is used as the target feature vector.

5. The welding parameter prediction method according to claim 4, wherein, The feature extraction module includes multiple gated loop units and a fully connected unit; The step of inputting the feature values ​​of the updated first welding parameters into the feature extraction module and obtaining the target hidden state output by the feature extraction module includes: For the first gated loop unit, the input data is the feature value after the first welding parameter is updated, and the output data is the hidden state of the first gated loop unit; For non-first gated loop units, the input data are the feature values ​​after the first welding parameter update and the hidden state output by the previous gated loop unit, and the output is the hidden state of the corresponding gated loop unit. The hidden state output by the last gated loop unit is input to the fully connected unit for full connection processing, and the hidden state output by the fully connected unit is obtained as the target hidden state.

6. The welding parameter prediction method according to claim 5, wherein, The second welding parameter prediction module includes multiple gated loop units and a global pooling unit; The step of inputting the target hiding state into the second welding parameter prediction module and outputting the feature value of the second welding parameter at the next moment includes: Determine the input data for each gated loop unit; wherein, for the first gated loop unit, the input data is the target hiding state; for non-first gated loop units, the input data is the hiding state output by the previous gated loop unit; For each gated loop unit, perform a dot product operation on the hidden state of its own output and output the predicted value of the second welding parameter; The predicted values ​​of the second welding parameters output by each gated loop unit are input into the global pooling unit. The predicted values ​​of all the second welding parameters are pooled to obtain the feature values ​​of the second welding parameters at the next time step.

7. The welding parameter prediction method according to any one of claims 1-6, wherein, The prediction model is generated in the following ways: The parameter values ​​of the first welding parameter and the second welding parameter at multiple time points are obtained in advance. The parameter values ​​of the first welding parameter and the second welding parameter at each time point are preprocessed to obtain the feature values ​​of the first welding parameter and the second welding parameter at each time point. The preprocessing includes at least one of data cleaning and normalization. Based on the characteristic values ​​of the first welding parameter and the second welding parameter at each time point, determine the characteristic values ​​of the first welding parameter at multiple sample times and the characteristic value of the second welding parameter at the next time point of the corresponding sample time. The initial prediction model is trained using the first welding parameter feature value at the sample time as the training sample and the second welding parameter feature value at the next time of the corresponding sample time as the training label, until the training stop condition is met, thus obtaining the welding prediction model.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory, the processor executing the computer program to implement the steps of the method according to any one of claims 1-7.

9. A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the method according to any one of claims 1-7.

Images