Welding method, device, equipment and medium

Abstract

The present application discloses a welding method, device, equipment, and medium, relating to the field of welding technology. The method comprises: obtaining the properties of the material to be welded; inputting the properties of the material to be welded into a preset network model; wherein the network model is obtained by training a data set consisting of property data for characterizing the welding material and welding parameters corresponding to each welding material, the data set comprising at least an empirical welding data set and an actual welding data set; outputting the welding parameters corresponding to the material to be welded through the network model; and sending the welding parameters to a welding system. It can be seen that welding of the material to be welded is achieved according to the welding parameters. Secondly, the data set comprises at least an empirical welding data set and an actual welding data set, and the property data and welding parameters in the data set are all multidimensional data. Therefore, the established network model is more robust, and the obtained welding parameters of the material to be welded are also more accurate.

Description

Technical Field

[0001] The present application relates to the field of welding technology, and in particular to a welding method, device, equipment and medium. Background Art

[0002] Resistance welding involves assembling the welded parts into overlapping joints and pressing them between two cylindrical electrodes. High current is used to generate resistance heat, locally heating the welded parts and melting the base metal, forming a nugget that cools to a weld spot. Spot welding is primarily used for welding thin sheets. It requires no filler metal, offers high productivity, minimizes weld deformation, and is easily automated, making it widely used in automobile manufacturing. However, precise resistance welding parameters are crucial during the welding process; otherwise, quality issues such as cold welds, cracks, deep indentations, and small welds can easily result. Substandard quality can lead to increased costs and even safety incidents if the vehicle is released to market.

[0003] It can be seen that how to select more accurate resistance spot welding process parameters and thus improve welding quality is a technical problem that people in this field urgently need to solve. Summary of the Invention

[0004] The purpose of this application is to provide a welding method, device, equipment and medium for obtaining more accurate resistance spot welding process parameters, so as to facilitate welding according to more accurate process parameters and improve welding quality.

[0005] To solve the above technical problems, the present application provides a welding method, comprising:

[0006] Get the properties of the material to be welded;

[0007] Inputting the properties of the current material to be welded into a preset network model; wherein the network model is obtained by training a data set consisting of property data for characterizing the welding material and welding parameters corresponding to each of the welding materials, the property data and the welding parameters being multidimensional data, and the data set comprising at least an empirical welding data set and an actual welding data set;

[0008] Outputting welding parameters corresponding to the current material to be welded through the network model;

[0009] The welding parameters are sent to a welding system so that the welding system welds the current material to be welded according to the welding parameters.

[0010] Preferably, establishing the network model includes:

[0011] obtaining the data set;

[0012] Preprocessing the data set and obtaining a preprocessed data set;

[0013] Dividing the preprocessed data set into training data, verification data, and test data according to a preset ratio;

[0014] Inputting the training data, the verification data, and the test data into the initial model for training respectively;

[0015] When the loss function of the test data is less than a threshold value, the weight corresponding to each welding parameter is obtained to complete the establishment of the network model.

[0016] Preferably, the preprocessing of the data set and obtaining the preprocessed data set includes:

[0017] Determining whether each of the attribute data and each of the welding parameters meets a corresponding first preset requirement;

[0018] If yes, then obtain the correlation coefficient between each property of the welding material and each welding parameter;

[0019] When the correlation coefficient meets the second preset requirement, normalizing each of the attribute data and each of the welding parameters to obtain the preprocessed data set;

[0020] If the correlation coefficient meets the second preset requirement, the attribute data that does not meet the second preset requirement is eliminated, and the step of normalizing each attribute data and each welding parameter to obtain the preprocessed data set is performed;

[0021] If not, the attribute data that do not meet the corresponding first preset requirements and the welding parameters corresponding to the attribute data are eliminated, and the process proceeds to the step of obtaining the correlation coefficients between the attributes of the welding material and the welding parameters.

[0022] Preferably, the determining whether each of the attribute data and each of the welding parameters meets the corresponding first preset requirement includes:

[0023] Selecting a target actual welding data set within a preset time period closest to current actual production from the actual welding data sets in the data sets;

[0024] Determining whether each of the attribute data and each of the welding parameters in the experience welding dataset and each of the target actual welding datasets meets the corresponding first preset requirements;

[0025] If so, proceed to the step of obtaining the correlation coefficients between the properties of the welding material and the welding parameters;

[0026] If not, the process proceeds to the step of eliminating the attribute data that does not meet the corresponding first preset requirement and the welding parameters corresponding to the attribute data.

[0027] Preferably, the initial model is a neural network model of a multilayer perceptron; wherein the number of neurons in the input layer is equal to the dimension of the remaining attribute data, and the remaining attribute data is the data obtained by removing the attribute data that does not meet the first preset requirement and the attribute data that does not meet the second preset requirement from the attribute data;

[0028] The input layer is connected to the hidden layer; the hidden layer is connected to the output layer;

[0029] The number of neurons in the output layer is equal to the dimension of the remaining welding parameters, and the remaining welding parameters are data obtained by removing the welding parameters that do not meet the first preset requirement from the welding parameters in the target actual welding data set;

[0030] The input layer and the hidden layer both contain activation functions and dropout functions;

[0031] The neural network model adopts back propagation and learning rate linear decay strategy.

[0032] Preferably, after outputting the welding parameters corresponding to the current material to be welded through the network model, the method further includes:

[0033] The data set is updated according to a preset frequency, and the process returns to the step of obtaining the data set.

[0034] Preferably, the initial model is at least any one of a neural network model, a multivariate linear regression model, a SVM model, and a decision tree model.

[0035] In order to solve the above technical problems, the present application also provides a welding device, comprising:

[0036] The acquisition module is used to obtain the properties of the current material to be welded;

[0037] An input module, configured to input the properties of the material to be welded into a preset network model; wherein the network model is obtained by training a data set consisting of property data for characterizing the welding material and welding parameters corresponding to each of the welding materials, wherein the property data and the welding parameters are multidimensional data, and the data set includes at least an empirical welding data set and an actual welding data set;

[0038] An output module, configured to output welding parameters corresponding to the current material to be welded through the network model;

[0039] The sending module is used to send the welding parameters to the welding system, so that the welding system welds the current material to be welded according to the welding parameters.

[0040] In order to solve the above technical problems, the present application also provides a welding device, comprising:

[0041] memory for storing computer programs;

[0042] A processor is used to implement the steps of the above-mentioned welding method when executing the computer program.

[0043] In order to solve the above technical problems, the present application also provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of the above welding method are implemented.

[0044] The welding method provided by the present application includes: obtaining the properties of the current material to be welded; inputting the properties of the current material to be welded into a preset network model; wherein the network model is obtained by training a data set consisting of property data for characterizing the welding material and welding parameters corresponding to each welding material, wherein the property data and welding parameters are multidimensional data, and the data set includes at least an empirical welding data set and an actual welding data set; outputting the welding parameters corresponding to the current material to be welded through the network model; and sending the welding parameters to a welding system so that the welding system welds the current material to be welded according to the welding parameters. As can be seen, in this method, welding of the material to be welded is achieved according to the welding parameters. Secondly, compared with a network model obtained by training a single data set and / or a data set that is all one-dimensional data, the method of the present application outputs the welding parameters corresponding to the material to be welded through the established network model. The data set in the method of the present application includes at least an empirical welding data set and an actual welding data set, and the property data and welding parameters in the data set are multidimensional data. Therefore, the established network model is more robust and the obtained welding parameters of the material to be welded are also more accurate.

[0045] In addition, the present application also provides a welding device, welding equipment and computer-readable storage medium, which have the same or corresponding technical features as the above-mentioned welding method and have the same effects as above. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] In order to more clearly illustrate the embodiments of the present application, the following is a brief introduction to the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0047] Figure 1A schematic diagram of the resistance welding structure provided in an embodiment of the present application;

[0048] Figure 2 A flow chart of a welding method provided in an embodiment of the present application;

[0049] Figure 3 A schematic diagram of a multi-layer perceptron neural network model provided in an embodiment of the present application;

[0050] Figure 4 A structural diagram of a welding device provided in one embodiment of the present application;

[0051] Figure 5 A structural diagram of a welding device provided in another embodiment of the present application;

[0052] Figure 6 A flowchart of a method for acquiring resistance welding process parameters based on self-learning provided in an embodiment of the present application. DETAILED DESCRIPTION

[0053] The following will be combined with the accompanying drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are only part of the embodiments of this application, not all of them. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0054] The core of this application is to provide a welding method, device, equipment and medium for obtaining more accurate resistance spot welding process parameters, so as to facilitate welding according to more accurate process parameters and improve welding quality.

[0055] Figure 1 This is a schematic diagram of a resistance welding structure provided in an embodiment of the present application. Figure 1 As shown, the structure at least comprises an upper electrode 1, a lower electrode 2, an object to be welded 3, a nugget 4, and a transformer 5. Resistance welding is a resistance welding method in which weldments are assembled into lap joints and pressed between two cylindrical electrodes, and high current is used to generate resistance heat to locally heat the weldments, melt the parent metal, form a nugget, and cool them into a weld. Spot welding is mainly used for thin plate welding. During welding, no filler metal is required, productivity is high, weldment deformation is small, automation is easy to achieve, and it is widely used in automobile manufacturing processes. During the welding process, it is necessary to set more accurate resistance welding parameters, otherwise it is easy to cause quality problems such as cold welding, cracks, too deep indentations, and too small welds. If the quality is unqualified, it will lead to increased costs, and even if the car enters the market, it may cause a safety accident. Therefore, in this application, a joint data set based on an expert knowledge base and real-time production process parameters of the factory is used to obtain more accurate resistance spot welding process parameters through a network model, thereby improving welding quality.

[0056] In order to enable those skilled in the art to better understand the present application, the present application is further described in detail below with reference to the accompanying drawings and specific implementation methods. Figure 2 A flow chart of a welding method provided in an embodiment of the present application is shown in FIG. Figure 2 As shown, the method includes:

[0057] S10: Obtain the properties of the material to be welded;

[0058] S11: Input the properties of the current material to be welded into the preset network model;

[0059] S12: Output welding parameters corresponding to the current material to be welded through the network model;

[0060] S13: Sending the welding parameters to the welding system so that the welding system can weld the current material to be welded according to the welding parameters.

[0061] To obtain the welding parameters corresponding to the material being welded, a network parameter model must first be established. In this embodiment, the network model is trained on a dataset consisting of attribute data representing the welding material and welding parameters corresponding to each welding material. Both the attribute data and the welding parameters are multidimensional data, and the datasets include at least an empirical welding dataset and an actual welding dataset.

[0062] Construct data source D (dataset) = expert dataset S (experienced welding dataset) + factory real-time production process parameter dataset F (actual welding dataset)

[0063] It should be noted that the data used to construct the data source are data obtained when the welding quality is normal.

[0064] D = {S, F} = [XS, YS];

[0065] Among them, XS represents a series of welding material properties, and YS represents the corresponding welding parameters.

[0066] The expert dataset S is the experience value given by the experts.

[0067] S = {S1, S2, ..., Si, ..., Sm}. In practice, m is not less than 1000.

[0068] Si=[{x1, x2, x3,…,xi,…,xp},{y1,y2,…,yi,…,yq}];

[0069] Among them, xi represents the comprehensive properties of the material, and its materials include DC01, DC03, DC06, HC340 / 590DP, Usibor1500P, etc. The material properties include p attribute values ​​such as thickness, density, resistivity, melting point, hardness, whether it is coated with glue, yield strength, and elongation after fracture. That is, the material properties are multi-dimensional data; yi represents the welding parameters, including q parameters such as current, pressure, and welding time. That is, the welding parameters are multi-dimensional data.

[0070] The factory real-time production process parameter data set F is also data obtained when the welding quality is normal.

[0071] F = {F1, F2, ..., Fi, ..., Fn}. In practice, n is not less than 1000.

[0072] Fi=[{x1,x2,x3,…,xi…,xp},{y1,y2,…,yi,…,yq}];

[0073] Similarly, xi represents the comprehensive properties of the material, including DC01, DC03, DC06, HC340 / 590DP, Usibor1500P, etc. The material properties include p attribute values ​​such as thickness, density, resistivity, melting point, hardness, whether it is coated with glue, yield strength, and elongation after fracture. yi represents welding parameters, including q parameters such as current, pressure, and welding time.

[0074] After acquiring the data set, it is input into the initial model for training to obtain a preset network model. Since the data set is multidimensional, it can be preprocessed, such as normalized, before entering the initial model. After obtaining the preset network model, the properties of the material to be welded are input into the preset network model, and the network model outputs the welding parameters corresponding to the material to be welded. After obtaining the welding parameters, manually inputting the data into the welding system is slow and prone to errors. Therefore, the parameters can be updated in the welding system by calling the welding system's application programming interface (API).

[0075] The welding method provided in this embodiment includes: obtaining the properties of the current material to be welded; inputting the properties of the current material to be welded into a preset network model; wherein the network model is obtained by training a data set consisting of property data for characterizing the welding material and welding parameters corresponding to each welding material, the property data and welding parameters are multi-dimensional data, and the data set includes at least an empirical welding data set and an actual welding data set; outputting the welding parameters corresponding to the current material to be welded through the network model; and sending the welding parameters to the welding system so that the welding system can weld the current material to be welded according to the welding parameters. It can be seen that in this method, welding of the material to be welded is achieved according to the welding parameters. Secondly, compared with the method of outputting the welding parameters corresponding to the material to be welded through the established network model, compared with the method of outputting the welding parameters corresponding to the material to be welded by training a single data set and / or the data sets are all one-dimensional data, the data set in the method of this embodiment includes at least an empirical welding data set and an actual welding data set, and the property data and welding parameters in the data set are all multi-dimensional data, so the established network model is more robust, and the obtained welding parameters of the material to be welded are also more accurate.

[0076] In order to obtain the welding parameters corresponding to the materials to be welded through the established network model, it is first necessary to establish a suitable network model. In practice, the preferred methods for establishing the network model include:

[0077] Get the dataset;

[0078] Preprocess the data set and obtain the preprocessed data set;

[0079] Divide the preprocessed data set into training data, verification data, and test data according to the preset ratio;

[0080] Input the training data, validation data, and test data into the initial model for training respectively;

[0081] When the loss function of the test data is less than the threshold, the weights corresponding to each welding parameter are obtained to complete the establishment of the network model.

[0082] Specifically, preprocessing the data set and obtaining the preprocessed data set includes:

[0083] Determining whether each attribute data and each welding parameter meets corresponding first preset requirements;

[0084] If so, the correlation coefficients between the properties of the welding material and the welding parameters are obtained;

[0085] When the correlation coefficient meets the second preset requirement, normalizing each attribute data and each welding parameter to obtain a preprocessed data set;

[0086] When the correlation coefficient meets the second preset requirement, the attribute data that does not meet the second preset requirement is eliminated, and the step of normalizing the attribute data and the welding parameters to obtain a preprocessed data set is performed;

[0087] If not, the attribute data that does not meet the corresponding first preset requirement and the welding parameters corresponding to the attribute data are eliminated, and the process proceeds to the step of obtaining the correlation coefficients between the various attributes of the welding material and the various welding parameters.

[0088] It should be noted that there is no limitation on the preset ratio, the first preset requirement, and the second preset requirement, and they are determined according to actual conditions. The initial model is at least any one of a neural network model, a multivariate linear regression model, a support vector machine (SVM) model, and a decision tree model. The above-mentioned process of judging whether each attribute data and each welding parameter meets the corresponding first preset requirement and eliminating the attribute data and the welding parameters corresponding to the attribute data that do not meet the corresponding first preset requirement can be considered as a process of eliminating outliers; judging whether the correlation coefficient meets the second preset requirement and eliminating the attribute data that do not meet the second preset requirement can be considered as a process of removing irrelevant factors, that is, the preprocessing process includes the process of eliminating outliers, removing irrelevant factors, and normalizing. Before eliminating data, in order to facilitate elimination of data, the various attributes of the material and the various welding parameters corresponding to the material can be arranged in a certain order, such as creating an EXCEL table and placing the data consisting of the various attributes of the material and the corresponding welding parameters in the same row.

[0089] Specifically, the steps to remove outliers are as follows:

[0090] Step 1: Eliminate data rows with custom data rules. The rules include greater than, less than, and interval. There is no limit on the set rules. They are determined according to the actual situation, such as the current is between [7K, 15K];

[0091] Step 2: Eliminate data rows outside of 3σ. If the data is not in (μ-3σ,μ+3σ), eliminate it. μ is the mean of the attribute dimension, and σ is the standard deviation of the attribute dimension.

[0092] When removing irrelevant factors, this embodiment selects dimensions by the Pearson correlation coefficient. Specifically,

[0093] The Pearson correlation coefficient formula is as follows:

[0094]

[0095] Where r is the Pearson correlation coefficient, ranging from -1 to 1, n is the number of data, X represents the welding material properties, and Y represents the welding parameters, which are selected from XS and YS respectively. are the means of the X and Y data, respectively. Calculate the correlation coefficients for different material properties against different welding parameters. If |r| for all y values ​​is less than a threshold (e.g., 0.1), then remove that dimension. For example, if the current value for each thickness is 7200A, then thickness does not affect the current value, so the thickness dimension can be removed.

[0096] After removing outliers and irrelevant factors, the processed dataset is normalized. Specifically, Z-score normalization is performed on each dimension of the processed D, where x' = (x - μ) / σ, where x' represents the normalized material property, μ is the mean, and σ is the standard deviation. After this preprocessing process, the resulting dataset is D'.

[0097] After obtaining the preprocessed dataset D', it is divided into training data, validation data, and test data according to a preset ratio. For example, the preprocessed dataset D' is randomly divided into training data, validation data, and test data in a ratio of 60%, 30%, and 10%. The training data, validation data, and test data are respectively input into the initial model for training. When the loss function of the test data is less than a threshold, the weights corresponding to each welding parameter are obtained to complete the network model. The threshold is not limited and is determined based on actual conditions.

[0098] The method of establishing a network model provided in this embodiment preprocesses the data by removing outliers, removing irrelevant factors and normalizing the data, and divides the preprocessed data into three different types of data: training data, verification data and test data, so as to improve the accuracy of the model.

[0099] In practice, in order to make the obtained data set more representative of data in actual production, a preferred embodiment is that determining whether each attribute data and each welding parameter meets the corresponding first preset requirement includes:

[0100] Selecting a target actual welding data set within a preset time period closest to current actual production from the actual welding data sets in the data set;

[0101] Determining whether each attribute data and each welding parameter in the experience welding data set and each target actual welding data set meet corresponding first preset requirements;

[0102] If so, proceed to the step of obtaining the correlation coefficients between various properties of the welding material and various welding parameters;

[0103] If not, the process proceeds to a step of eliminating the attribute data that does not meet the corresponding first preset requirement and the welding parameters corresponding to the attribute data.

[0104] The initial data set contains an actual welding data set. However, this actual welding data set may be data from the actual welding process in the past, such as data from two years ago. However, with the development of science and technology, the properties of the materials may have changed greatly. If the data from two years ago is still used, the resulting network model may not be robust. Therefore, in this embodiment, the actual welding data of the most recent period is selected. There is no limit on the value of the preset time length. This process can be called time window interception (only for factory process parameters). The production data FA = {F1, F2, ..., Fi, ..., Fr} within the most recent period (such as 3 months) is selected from F, where r represents the number of data items in the time window, and F = FA. After the time window interception is performed, the step of removing outliers in the above embodiment can be entered.

[0105] This embodiment provides the method of selecting production data from a recent period of time to establish a network model, so that the data is more referenceable and the accuracy of network model prediction can be improved.

[0106] When establishing the network model, the initial model selected is a multi-layer perceptron neural network model; wherein the number of neurons in the input layer is equal to the dimension of the remaining attribute data, and the remaining attribute data is the data obtained by eliminating the attribute data that does not meet the first preset requirement and the attribute data that does not meet the second preset requirement from the attribute data;

[0107] The input layer is connected to the hidden layer; the hidden layer is connected to the output layer;

[0108] The number of neurons in the output layer is equal to the dimension of the remaining welding parameters, and the remaining welding parameters are the data obtained by eliminating the welding parameters that do not meet the first preset requirement from the welding parameters in the target actual welding data set;

[0109] The input layer and hidden layer both contain activation functions and dropout functions;

[0110] The neural network model adopts back propagation and learning rate linear decay strategy.

[0111] Construct a multi-layer perceptron neural network model, including the number of neurons in the input layer (the number is equal to the characteristic dimension of XS obtained after elimination), the number of neurons in the hidden layer (greater than or equal to 1), and the number of neurons in the output layer (the number is equal to the characteristic dimension of YS obtained after elimination). Figure 3 This is a schematic diagram of a multi-layer perceptron neural network model provided in the embodiment of the present application. Figure 3 As shown in Figure 1, the model includes an input layer, a hidden layer, and an output layer. The input layer and the hidden layer contain a LeakyReLU layer and a dropout layer respectively.

[0112] Among them, the activation function uses LeakyReLU, α = 0.1, and the calculation formula of the activation function LeakyReLU is as follows:

[0113]

[0114] To prevent overfitting, a dropout strategy is added, where r is a random value between 0 and 1, p is the dropout probability, and the dropout probability is set to 0.5. The dropout calculation formula is as follows:

[0115]

[0116] The initialization weights are initialized using a standard normal distribution with a mean of 0 and a standard deviation of 1.

[0117] Loss is a weighted MSE loss, where loss = Ew(y^hat - y)², where y^hat is the predicted value, y is the true value, and w is the category weight. You can assign larger weights to important parameters like current and pressure, while assigning smaller weights to less important parameters like welding time and preheating time.

[0118] For back propagation, the optimizer can be Stochastic Gradient Descent (SGD), Root Mean Square Prop (RMSprop), Adaptive Gradient (Adagrad), or Adaptive Moment Estimation (Adam).

[0119] Establish a learning rate linear decay strategy lr = γ * lr, the initial learning rate lr = 0.01, and γ is 0.1

[0120] There is no limit on the iteration cycle, for example, the iteration cycle is selected to be greater than 1000 rounds.

[0121] During training, when the test data loss is less than the set threshold, training is stopped and the weights are saved to meet the online requirements.

[0122] It should be noted that, in addition to the multi-layer perceptron model used in this embodiment, other machine learning models may also be used in practice, such as a multiple linear regression model, an SVM model, a decision tree, and the like.

[0123] As time goes by, the influence of external uncontrollable factors, such as changes in plate material, changes in plate quality uniformity, changes in plate contact angle, plate cleanliness, electrode alignment, changes in electrode material, changes in electrode shape, aging of welding guns, etc., leads to changes in the welding environment, resulting in increasingly poor welding results. Therefore, it is necessary to update the model. In a preferred embodiment, after outputting the welding parameters corresponding to the current material to be welded through the network model, the welding method further includes:

[0124] Update the dataset at the preset frequency and return to the step of obtaining the dataset.

[0125] There is no limit on the preset frequency and it is determined based on actual conditions. According to the set update cycle (day, week, or month), the production data in D is replaced and preprocessed to obtain the standard data set D".

[0126] The weights of the first established network model are used as the initial weights, some layers are frozen, a small learning rate lr is set, such as 0.001, and multiple rounds of training are performed, such as 1000 rounds. It should be noted that when freezing some layers, the number of frozen layers should not be too large, otherwise the previous experience values ​​will be overwritten. If the number of frozen layers is too small, the latest data cannot be trained as much. Therefore, it is necessary to select an appropriate number of frozen layers.

[0127] Establish an early stopping strategy. When the test loss is less than a predetermined threshold, terminate the current training. Replace the previous inference model with the optimal inference model.

[0128] This embodiment provides a self-learning update model so that the obtained welding parameters are more accurate, so that welding can be performed according to the more accurate welding parameters, thereby improving the quality of welding.

[0129] In the above embodiments, the welding method is described in detail. This application also provides corresponding embodiments of the welding device and welding equipment. It should be noted that this application describes the embodiments of the device part from two perspectives: one is based on the functional module perspective, and the other is based on the hardware perspective.

[0130] Figure 4 This is a structural diagram of a welding device provided in one embodiment of the present application. This embodiment is based on the perspective of functional modules and includes:

[0131] An acquisition module 10 is used to obtain the properties of the material to be welded;

[0132] An input module 11 is used to input the properties of the material to be welded into a preset network model; wherein the network model is obtained by training a data set consisting of property data for characterizing the welding material and welding parameters corresponding to each welding material, wherein the property data and welding parameters are multidimensional data, and the data set includes at least an empirical welding data set and an actual welding data set;

[0133] Output module 12, used to output welding parameters corresponding to the current material to be welded through the network model;

[0134] The sending module 13 is used to send the welding parameters to the welding system so that the welding system can weld the current material to be welded according to the welding parameters.

[0135] Since the embodiments of the device part correspond to the embodiments of the method part, the embodiments of the device part refer to the description of the embodiments of the method part, which will not be described here. The device part has the same beneficial effects as the welding method mentioned above.

[0136] Figure 5 This is a structural diagram of a welding device provided in another embodiment of the present application. This embodiment is based on the hardware perspective, such as Figure 5 As shown, the welding equipment includes:

[0137] Memory 20, for storing computer programs;

[0138] The processor 21 is configured to implement the steps of the welding method mentioned in the above embodiment when executing the computer program.

[0139] Among them, the processor 21 may include one or more processing cores, such as a 4-core processor, an 8-core processor, etc. The processor 21 can be implemented in at least one hardware form of a digital signal processor (DSP), a field programmable gate array (FPGA), and a programmable logic array (PLA). The processor 21 may also include a main processor and a coprocessor. The main processor is a processor for processing data in the awake state, also known as a central processing unit (CPU); the coprocessor is a low-power processor for processing data in the standby state. In some embodiments, the processor 21 may be integrated with a graphics processing unit (GPU), which is responsible for rendering and drawing the content to be displayed on the display screen. In some embodiments, the processor 21 may also include an artificial intelligence (AI) processor, which is used to process computing operations related to machine learning.

[0140] The memory 20 may include one or more computer-readable storage media, which may be non-transitory. The memory 20 may also include high-speed random access memory, and non-volatile memory, such as one or more disk storage devices, flash memory storage devices. In this embodiment, the memory 20 is at least used to store the following computer program 201, wherein, after the computer program is loaded and executed by the processor 21, it can implement the relevant steps of the welding method disclosed in any of the aforementioned embodiments. In addition, the resources stored in the memory 20 may also include an operating system 202 and data 203, etc., and the storage method may be temporary storage or permanent storage. Among them, the operating system 202 may include Windows, Unix, Linux, etc. The data 203 may include but is not limited to the data involved in the welding method mentioned above.

[0141] In some embodiments, the welding device may further include a display screen 22 , an input / output interface 23 , a communication interface 24 , a power source 25 , and a communication bus 26 .

[0142] Those skilled in the art will understand that Figure 5 The structure shown in the figure does not constitute a limitation of the welding device, and may include more or fewer components than shown.

[0143] The welding device provided in the embodiment of the present application includes a memory and a processor. When the processor executes the program stored in the memory, it can implement the following method: welding method, the effect is the same as above.

[0144] The present application also provides an embodiment corresponding to a computer-readable storage medium. The computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps described in the above method embodiment.

[0145] It is understandable that if the method in the above embodiment is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product, which is stored in a storage medium and executes all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0146] The computer-readable storage medium provided in this application includes the welding method mentioned above, and the effect is the same as above.

[0147] In order to enable those skilled in the art to better understand the present application, the present application is further described in detail below with reference to the accompanying drawings and specific implementation methods. Figure 6 Flowchart of the method for obtaining resistance welding process parameters based on self-learning provided in the embodiment of the present application. Figure 6 As shown, the method includes:

[0148] S14: Construct a joint dataset based on expert knowledge base and factory real-time production industrial parameters;

[0149] S15: Preprocessing uses threshold method and three sigma to eliminate outliers, and Pearson correlation coefficient to screen features;

[0150] S16: Build a neural network model based on a multi-layer perceptron and practice back-propagation to optimize model parameters;

[0151] S17: Use linear decay learning rate and early stopping mechanism to build a self-learning system;

[0152] S18: The predicted welding process parameters are set into the welding system through the API interface.

[0153] In the method provided in this embodiment, the basic data is expert experience data + all real-time production welding data. A single model is applicable to all production plates, which can solve the problem of a small scope of application. Select the data time window (3 months), select which dimensions (correlation analysis>0.1), and how many to eliminate (threshold, 3σ). Eliminate interference to solve the problem of inaccurate recommended parameters; add LeakReLU activation function, dropout to prevent overfitting, etc. to ensure the effect and solve the problem of inaccurate recommended parameters; dynamically update the model. Dynamically fit the actual situation on site according to the production data to solve the problem that the recommended parameters become worse over time; update the parameters to the equipment by calling the API interface of the welding equipment, solving the problem of slow and error-prone manual entry.

[0154] The above is a detailed introduction to a welding method, device, equipment and medium provided by the present application. The various embodiments in the specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same and similar parts between the various embodiments can be referred to each other. For the device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part description. It should be pointed out that for ordinary technicians in this technical field, without departing from the principles of this application, several improvements and modifications can be made to the present application, and these improvements and modifications also fall within the scope of protection of the claims of this application.

[0155] It should also be noted that, in this specification, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variants thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus comprising a series of elements includes not only those elements, but also other elements not explicitly listed, or elements inherent to such process, method, article, or apparatus. In the absence of further limitations, an element defined by the phrase "comprising a ..." does not exclude the presence of additional identical elements in the process, method, article, or apparatus comprising the element.

Claims

1. A welding method, characterized in that: include: Get the properties of the material to be welded; Inputting the properties of the current material to be welded into a preset network model; wherein the network model is obtained by training a data set consisting of property data for characterizing the welding material and welding parameters corresponding to each of the welding materials, the property data and the welding parameters being multidimensional data, and the data set comprising at least an empirical welding data set and an actual welding data set; Outputting welding parameters corresponding to the current material to be welded through the network model; Sending the welding parameters to a welding system so that the welding system welds the current material to be welded according to the welding parameters; Wherein, establishing the network model includes: obtaining the data set; Preprocessing the data set and obtaining a preprocessed data set, wherein the preprocessed data set is used to establish the network model; Preprocessing the data set and obtaining the preprocessed data set includes: Determining whether each of the attribute data and each of the welding parameters meets a corresponding first preset requirement; If yes, then obtain the correlation coefficient between each property of the welding material and each welding parameter; When the correlation coefficient meets the second preset requirement, normalizing each of the attribute data and each of the welding parameters to obtain the preprocessed data set; If the correlation coefficient meets the second preset requirement, the attribute data that does not meet the second preset requirement is eliminated, and the step of normalizing each attribute data and each welding parameter to obtain the preprocessed data set is performed; If not, the attribute data and the welding parameters corresponding to the attribute data that do not meet the corresponding first preset requirements are eliminated, and the step of obtaining the correlation coefficient between each attribute of the welding material and each welding parameter is entered; The determining whether each of the attribute data and each of the welding parameters meets the corresponding first preset requirements includes: Selecting a target actual welding data set within a preset time period closest to current actual production from the actual welding data sets in the data sets; Determining whether each of the attribute data and each of the welding parameters in the experience welding dataset and each of the target actual welding datasets meets the corresponding first preset requirements; If so, proceed to the step of obtaining the correlation coefficients between the properties of the welding material and the welding parameters; If not, the process proceeds to the step of eliminating the attribute data that does not meet the corresponding first preset requirement and the welding parameters corresponding to the attribute data.

2. The welding method according to claim 1, characterized in that Establishing the network model further includes: dividing the preprocessed data set into training data, verification data, and test data according to a preset ratio; Inputting the training data, the verification data, and the test data into the initial model for training respectively; When the loss function of the test data is less than a threshold value, the weight corresponding to each welding parameter is obtained to complete the establishment of the network model.

3. The welding method according to claim 2, characterized in that The initial model is a neural network model of a multilayer perceptron; wherein the number of neurons in the input layer is equal to the dimension of the remaining attribute data, and the remaining attribute data is the data obtained by removing the attribute data that does not meet the first preset requirement and the attribute data that does not meet the second preset requirement from the attribute data; The input layer is connected to the hidden layer; the hidden layer is connected to the output layer; The number of neurons in the output layer is equal to the dimension of the remaining welding parameters, and the remaining welding parameters are data obtained by removing the welding parameters that do not meet the first preset requirement from the welding parameters in the target actual welding data set; The input layer and the hidden layer both contain activation functions and dropout functions; The neural network model adopts back propagation and learning rate linear decay strategy.

4. The welding method according to any one of claims 2 to 3, characterized in that: After outputting the welding parameters corresponding to the current material to be welded through the network model, the method further includes: The data set is updated according to a preset frequency, and the process returns to the step of obtaining the data set.

5. The welding method according to claim 2, characterized in that The initial model is at least any one of a neural network model, a multivariate linear regression model, a SVM model, and a decision tree model.

6. A welding device, characterized in that: include: The acquisition module is used to obtain the properties of the current material to be welded; An input module is used to input the properties of the current material to be welded into a preset network model; wherein the network model is obtained by training a data set consisting of property data for characterizing the welding material and welding parameters corresponding to each welding material, wherein the property data and the welding parameters are multidimensional data, and the data set includes at least an empirical welding data set and an actual welding data set; wherein establishing the network model includes: obtaining the data set; Preprocessing the data set and obtaining a preprocessed data set, wherein the preprocessed data set is used to establish the network model; Preprocessing the data set and obtaining the preprocessed data set includes: Determining whether each of the attribute data and each of the welding parameters meets a corresponding first preset requirement; If yes, then obtain the correlation coefficient between each property of the welding material and each welding parameter; When the correlation coefficient meets the second preset requirement, normalizing each of the attribute data and each of the welding parameters to obtain the preprocessed data set; If the correlation coefficient meets the second preset requirement, the attribute data that does not meet the second preset requirement is eliminated, and the step of normalizing each attribute data and each welding parameter to obtain the preprocessed data set is performed; If not, the attribute data and the welding parameters corresponding to the attribute data that do not meet the corresponding first preset requirements are eliminated, and the step of obtaining the correlation coefficient between each attribute of the welding material and each welding parameter is entered; The determining whether each of the attribute data and each of the welding parameters meets the corresponding first preset requirements includes: Selecting a target actual welding data set within a preset time period closest to current actual production from the actual welding data sets in the data sets; Determining whether each of the attribute data and each of the welding parameters in the experience welding dataset and each of the target actual welding datasets meets the corresponding first preset requirements; If so, proceed to the step of obtaining the correlation coefficients between the properties of the welding material and the welding parameters; If not, the step of eliminating the attribute data that does not meet the corresponding first preset requirement and the welding parameters corresponding to the attribute data is entered; An output module, configured to output welding parameters corresponding to the current material to be welded through the network model; The sending module is used to send the welding parameters to the welding system, so that the welding system welds the current material to be welded according to the welding parameters.

7. A welding device, characterized in that: include: Memory for storing computer programs; A processor, configured to implement the steps of the welding method according to any one of claims 1 to 5 when executing the computer program.

8. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the welding method according to any one of claims 1 to 5 are implemented.

Images