Method, device, medium and terminal for predicting stress and strain of ingot during continuous casting

Abstract

The present application discloses a method and device, medium and terminal for predicting the stress and strain of the ingot during the continuous casting process, which relates to the field of continuous casting technology. The main purpose is to improve the existing problem of huge workload caused by the deformation mechanism of the material in different directions and temperatures, which requires scene-by-scene modeling, and the problem of greatly increased experimental costs caused by the lack of key data points in the fitting process, which requires a large amount of experimental data. It includes: obtaining the continuous casting process parameters and casting machine parameters for the target material; based on the ingot strain and stress prediction model that has completed model training, according to the continuous casting process parameters and casting machine parameters, predicting the ingot strain and stress distribution data of the target material, the ingot strain and stress prediction model is constructed based on a conditional generative adversarial network, and is obtained by performing two model trainings in turn using a preset single material simulation continuous casting data set and the experimental continuous casting data set of the target material.

Description

Technical Field

[0001] The present application relates to the field of continuous casting technology, and in particular to a method and device, medium, and terminal for predicting stress and strain of a casting during a continuous casting process. Background Art

[0002] Continuous casting billet is a product obtained by casting molten steel from a steelmaking furnace through a continuous casting machine. During the continuous casting process, if the comprehensive strain at the solidification front of the billet exceeds its critical strain, it will cause internal cracks in the billet, thereby reducing product quality. Reproduction will increase production costs and reduce production efficiency. Therefore, accurately predicting the deformation behavior of continuous casting billets during the production process is of great significance for improving production quality and efficiency.

[0003] In order to accurately predict the stress field during solidification, it is necessary to establish an accurate mechanical constitutive model based on the high-temperature mechanical behavior of the material. The traditional stress-strain constitutive model usually uses the Arrhenius model to describe the relationship between the deformation conditions and the rheological stress of the material during thermal deformation. The Arrhenius model uses a relatively fixed fitting process to solve the relationship between stress and temperature, strain rate and strain.

[0004] However, due to the deformation mechanism of materials in different directions and temperatures, it is necessary to model the scenes separately, which is a huge workload. In addition, since the loss of key data points is not allowed in the fitting process, a large amount of experimental data is required, which greatly increases the experimental cost. Therefore, a method for predicting the stress and strain of the ingot during continuous casting is urgently needed to improve the above problems. Summary of the invention

[0005] In view of this, the present application provides a method and device, medium, and terminal for predicting the stress and strain of the ingot during the continuous casting process. The main purpose is to improve the existing problem of huge workload caused by the deformation mechanism of the material in different directions and temperatures, which requires scene-by-scene modeling, and the problem of greatly increased experimental costs due to the fact that the loss of key data points is not allowed in the fitting process and a large amount of experimental data is required.

[0006] According to one aspect of the present application, a method for predicting stress and strain of a casting during a continuous casting process is provided, comprising:

[0007] Obtain continuous casting process parameters and casting machine parameters for target materials;

[0008] Based on the ingot strain and stress prediction model that has completed model training, the ingot strain and stress distribution data of the target material is predicted according to the continuous casting process parameters and the casting machine parameters. The ingot strain and stress prediction model is constructed based on a conditional generative adversarial network, and is trained twice using a preset single material simulation continuous casting data set and an experimental continuous casting data set of the target material in sequence. The loss function used in the model training is corrected by the numerical deviation and curve morphology deviation between the predicted ingot strain and stress distribution data and the simulated ingot strain and stress distribution data.

[0009] Preferably, the cast billet strain stress prediction model includes a generator and a discriminator;

[0010] The generator is used to generate predicted strain stress distribution data of the ingot according to the continuous casting process parameters and the casting machine parameters;

[0011] The discriminator is used to discriminate the authenticity of the predicted slab strain and stress distribution data according to the simulated slab strain and stress distribution data in the simulated continuous casting data set;

[0012] Based on the strain and stress prediction model of the cast billet that has completed model training, according to the continuous casting process parameters and the casting machine parameters, the strain and stress distribution data of the cast billet of the target material is predicted, including:

[0013] The generator in the slab strain and stress prediction model based on the completed model training predicts the slab strain and stress distribution data of the target material according to the continuous casting process parameters and the casting machine parameters.

[0014] Preferably, before the strain and stress prediction model for casting billet based on the completed model training predicts the strain and stress distribution data of the casting billet of the target material according to the continuous casting process parameters and the casting machine parameters, the method further comprises:

[0015] Constructing an initial generator based on a long short-term memory network, constructing an initial discriminator based on a convolutional neural network, and combining the initial generator and the initial discriminator to construct an initial casting billet strain stress prediction model;

[0016] Acquire a simulated continuous casting data set of a preset single material, wherein the simulated continuous casting data set includes a first continuous casting process parameter, a first casting machine parameter, and a first simulated cast billet strain stress distribution data for the preset single material;

[0017] Based on the initial generator, according to the first continuous casting process parameters and the first casting machine parameters, predict and obtain first predicted cast billet strain stress distribution data of the preset single material;

[0018] Based on the initial discriminator, according to the first continuous casting process parameters and the first casting machine parameters, the first simulated ingot strain and stress distribution data and the first predicted ingot strain and stress distribution data are distinguished to obtain a first authenticity discrimination result;

[0019] Calculating a first loss function between the first authenticity discrimination result and the authenticity result;

[0020] Update the model parameters of the initial generator and the initial discriminator based on the first loss function to obtain an intermediate generator and an intermediate discriminator;

[0021] Acquire an experimental continuous casting data set of the target material, wherein the experimental continuous casting data set includes a second continuous casting process parameter, a second casting machine parameter, and experimental cast billet strain stress distribution data for the target material;

[0022] Based on the intermediate generator, according to the second continuous casting process parameters and the second casting machine parameters, predict and obtain second predicted cast slab strain stress distribution data of the target material;

[0023] Based on the intermediate discriminator, according to the second continuous casting process parameters and the second casting machine parameters, the experimental ingot strain and stress distribution data and the second predicted ingot strain and stress distribution data are distinguished to obtain a second authenticity discrimination result;

[0024] Calculating a second loss function between the second authenticity discrimination result and the authenticity result;

[0025] Based on the second loss function, the model parameters of the intermediate generator and the intermediate discriminator are updated to obtain the generator and the discriminator that have completed model training, so as to predict the strain and stress distribution data of the ingot based on the continuous casting process parameters and the casting machine parameters based on the generator that has completed model training.

[0026] Preferably, before updating the model parameters of the initial generator and the initial discriminator based on the first loss function to obtain the intermediate generator and the intermediate discriminator, the method further includes:

[0027] The first numerical deviation between the first predicted cast billet strain stress distribution data and the first simulated cast billet strain stress distribution data is calculated according to the following formula:

[0028]

[0029] in, Indicates the numerical deviation. Indicates the first predicted slab strain stress distribution data elements, The first simulation shows the strain stress distribution data of the first slab. elements, n represents the number of cross-sectional strain distribution points;

[0030] The first curve shape deviation between the first predicted cast billet strain stress distribution data and the first simulated cast billet strain stress distribution data is calculated according to the following formula:

[0031]

[0032] in, Indicates the curve shape deviation, The first simulation shows the difference component d of the strain stress distribution data of the first simulation casting. The value of the dimension, Represents the first predicted slab strain stress distribution data difference component No. The value of the dimension;

[0033] Correcting the first loss function according to the first numerical deviation and the first curve shape deviation to obtain a corrected first loss function, and updating the model parameters of the initial generator and the initial discriminator based on the corrected first loss function;

[0034] And before updating the model parameters of the intermediate generator and the intermediate discriminator based on the second loss function to obtain the generator and the discriminator that have completed model training, the method also includes:

[0035] According to the above formula, a second numerical deviation between the second predicted cast billet strain stress distribution data and the experimental cast billet strain stress distribution data is calculated;

[0036] According to the above formula, a second curve shape deviation between the second predicted cast billet strain stress distribution data and the experimental cast billet strain stress distribution data is calculated;

[0037] The second loss function is corrected according to the second numerical deviation and the second curve morphology deviation to obtain a corrected second loss function, so as to update the model parameters of the intermediate generator and the intermediate discriminator based on the corrected second loss function.

[0038] Preferably, before predicting and obtaining the first predicted cast billet strain stress distribution data of the preset single material based on the initial generator and according to the first continuous casting process parameters and the first casting machine parameters, the method further comprises:

[0039] Performing Z-Score standardization on the data contained in the simulated continuous casting data set so that the mean of each data is zero and the variance is 1, so as to perform prediction processing based on the standardized simulated continuous casting data set based on the initial generator;

[0040] Before predicting and obtaining the second predicted cast slab strain stress distribution data of the target material based on the intermediate generator and according to the second continuous casting process parameters and the second casting machine parameters, the method further includes:

[0041] The data contained in the experimental continuous casting data set is subjected to Z-Score standardization processing so that the mean of each data is zero and the variance is 1, so as to perform prediction processing based on the standardized experimental continuous casting data set based on the intermediate generator.

[0042] Preferably, the step of obtaining a simulated continuous casting data set of a preset single material includes:

[0043] Acquire a first casting machine parameter and a first continuous casting process parameter of the preset single material;

[0044] Establishing a geometric model of the ingot based on the first casting machine parameters;

[0045] Based on the first continuous casting process parameters and the geometric model, a simulation operation is performed using finite element software to obtain a simulation continuous casting data set of the preset single material.

[0046] Preferably, the continuous casting process parameters include material composition, solid phase ratio, reduction amount, and casting speed.

[0047] According to another aspect of the present application, a device for predicting stress and strain of a casting during a continuous casting process is provided, comprising:

[0048] A parameter acquisition module is used to obtain continuous casting process parameters and casting machine parameters for target materials;

[0049] A strain and stress prediction module is used to predict the strain and stress distribution data of the target material based on the strain and stress prediction model of the ingot that has completed model training, according to the continuous casting process parameters and the casting machine parameters. The strain and stress prediction model of the ingot is constructed based on a conditional generative adversarial network, and is obtained by performing two model trainings in sequence using a simulated continuous casting data set of a preset single material and an experimental continuous casting data set of the target material. The loss function used in the model training is corrected by the numerical deviation and curve morphology deviation between the predicted strain and stress distribution data of the ingot and the simulated strain and stress distribution data of the ingot.

[0050] Preferably, the cast billet strain stress prediction model includes a generator and a discriminator;

[0051] The generator is used to generate predicted strain stress distribution data of the ingot according to the continuous casting process parameters and the casting machine parameters;

[0052] The discriminator is used to discriminate the authenticity of the predicted slab strain and stress distribution data according to the simulated slab strain and stress distribution data in the simulated continuous casting data set;

[0053] The strain stress prediction module is used to:

[0054] Based on the generator in the slab strain and stress prediction model for which model training has been completed, the slab strain and stress distribution data of the target material is predicted according to the continuous casting process parameters and the casting machine parameters.

[0055] Preferably, before the strain stress prediction module, the device further includes a model training module, including:

[0056] A model building unit, used to build an initial generator based on a long short-term memory network, build an initial discriminator based on a convolutional neural network, and combine the initial generator and the initial discriminator to build an initial casting billet strain stress prediction model;

[0057] A simulated continuous casting data set acquisition unit, used to acquire a simulated continuous casting data set of a preset single material, wherein the simulated continuous casting data set includes a first continuous casting process parameter, a first casting machine parameter, and a first simulated cast billet strain stress distribution data for the preset single material;

[0058] A first prediction unit is used to predict and obtain first predicted cast slab strain stress distribution data of the preset single material based on the initial generator, according to the first continuous casting process parameters and the first casting machine parameters;

[0059] A first discriminating unit is configured to discriminate the first simulated ingot strain and stress distribution data and the first predicted ingot strain and stress distribution data based on the initial discriminator and the first continuous casting process parameter and the first casting machine parameter to obtain a first authenticity discrimination result;

[0060] A first loss function calculation unit, used to calculate a first loss function between the first authenticity discrimination result and the authenticity result;

[0061] A first parameter updating unit, used for updating the model parameters of the initial generator and the initial discriminator based on the first loss function to obtain an intermediate generator and an intermediate discriminator;

[0062] An experimental continuous casting data set acquisition unit, used to acquire an experimental continuous casting data set of the target material, wherein the experimental continuous casting data set includes a second continuous casting process parameter and a second casting machine parameter for the target material, and experimental cast billet strain stress distribution data;

[0063] A second prediction unit is used to predict and obtain second predicted cast slab strain stress distribution data of the target material based on the intermediate generator, according to the second continuous casting process parameters and the second casting machine parameters;

[0064] A second discrimination unit is used to distinguish the experimental ingot strain and stress distribution data and the second predicted ingot strain and stress distribution data based on the intermediate discriminator and according to the second continuous casting process parameters and the second casting machine parameters, so as to obtain a second authenticity discrimination result;

[0065] A second loss function calculation unit, used to calculate a second loss function between the second authenticity discrimination result and the authenticity result;

[0066] The second parameter updating unit is used to update the model parameters of the intermediate generator and the intermediate discriminator based on the second loss function to obtain the generator and the discriminator that have completed the model training, so as to predict the strain and stress distribution data of the ingot based on the continuous casting process parameters and the casting machine parameters of the generator that has completed the model training.

[0067] Preferably, before the first parameter updating unit, the model training module further includes a first loss function correction unit, which is used to:

[0068] The first numerical deviation between the first predicted cast billet strain stress distribution data and the first simulated cast billet strain stress distribution data is calculated according to the following formula:

[0069]

[0070] in, Indicates the numerical deviation. Indicates the first predicted slab strain stress distribution data elements, The first simulation shows the strain stress distribution data of the first slab. elements, n represents the number of cross-sectional strain distribution points;

[0071] The first curve shape deviation between the first predicted cast billet strain stress distribution data and the first simulated cast billet strain stress distribution data is calculated according to the following formula:

[0072]

[0073] in, Indicates the curve shape deviation, The first simulation shows the difference component d of the strain stress distribution data of the first simulation casting. The value of the dimension, Represents the first prediction of the difference component of the strain stress distribution data of the slab No. The value of the dimension;

[0074] Correcting the first loss function according to the first numerical deviation and the first curve shape deviation to obtain a corrected first loss function, and updating the model parameters of the initial generator and the initial discriminator based on the corrected first loss function;

[0075] And before the second parameter updating unit, the model training module further includes a second loss function correction unit, which is used to:

[0076] According to the above formula, a second numerical deviation between the second predicted cast billet strain stress distribution data and the experimental cast billet strain stress distribution data is calculated;

[0077] According to the above formula, a second curve shape deviation between the second predicted cast billet strain stress distribution data and the experimental cast billet strain stress distribution data is calculated;

[0078] The second loss function is corrected according to the second numerical deviation and the second curve morphology deviation to obtain a corrected second loss function, so as to update the model parameters of the intermediate generator and the intermediate discriminator based on the corrected second loss function.

[0079] Preferably, before the first prediction unit, the strain stress prediction module further includes:

[0080] A first standardization processing unit is used to perform Z-Score standardization processing on the data contained in the simulated continuous casting data set so that the mean of each data is zero and the variance is 1, so as to perform prediction processing based on the simulated continuous casting data set after the standardization processing based on the initial generator;

[0081] Before the second prediction unit, the strain stress prediction module further includes:

[0082] The second standardization processing unit is used to perform Z-Score standardization on the data contained in the experimental continuous casting data set so that the mean of each data is zero and the variance is 1, so as to perform prediction processing based on the standardized experimental continuous casting data set based on the intermediate generator.

[0083] Preferably, the simulation continuous casting data set acquisition unit is used to:

[0084] Acquire a first casting machine parameter and a first continuous casting process parameter of the preset single material;

[0085] Establishing a geometric model of the ingot based on the first casting machine parameters;

[0086] Based on the first continuous casting process parameters and the geometric model, a simulation operation is performed using finite element software to obtain a simulation continuous casting data set of the preset single material.

[0087] Preferably, the continuous casting process parameters include material composition, solid phase ratio, reduction amount, and casting speed.

[0088] According to another aspect of the present application, a storage medium is provided, wherein at least one executable instruction is stored in the storage medium, and the executable instruction enables a processor to execute operations corresponding to the above-mentioned method for predicting stress and strain of a cast billet during a continuous casting process.

[0089] According to another aspect of the present application, a terminal is provided, comprising: a processor, a memory, a communication interface and a communication bus, wherein the processor, the memory and the communication interface communicate with each other through the communication bus;

[0090] The memory is used to store at least one executable instruction, and the executable instruction enables the processor to execute operations corresponding to the above-mentioned method for predicting stress and strain of the cast billet during the continuous casting process.

[0091] By means of the above technical solution, the technical solution provided by the embodiment of the present application has at least the following advantages:

[0092] The present application provides a method and device, medium, and terminal for predicting stress and strain of a cast billet during a continuous casting process. First, continuous casting process parameters and casting machine parameters for a target material are obtained. Furthermore, based on a cast billet strain and stress prediction model for which model training has been completed, cast billet strain and stress distribution data of the target material are predicted according to the continuous casting process parameters and the casting machine parameters. The cast billet strain and stress prediction model is constructed based on a conditional generative adversarial network, and is obtained by performing two model trainings in sequence using a simulated continuous casting data set of a preset single material and an experimental continuous casting data set of the target material. The loss function used in the model training is obtained by correcting the numerical deviation and curve morphology deviation between the predicted cast billet strain and stress distribution data and the simulated cast billet strain and stress distribution data. Compared with the prior art, the embodiment of the present application constructs a strain and stress prediction model for the ingot based on a conditional generative adversarial network, and uses the real data obtained by simulation and the generated prediction data for adversarial training. Since the adversarial training is essentially based on the generator to generate predicted ingot strain and stress distribution data, and then based on the discriminator to discriminate the authenticity of the predicted ingot strain and stress distribution data, by training the generator and the discriminator at the same time, the generator continuously improves the fidelity of the prediction data, and the discriminator continuously improves the ability to distinguish true and false data, thereby improving the accuracy and stability of the ingot strain and stress prediction model, thereby improving the prediction accuracy of the ingot stress and strain in the continuous casting process; further, using A simulated continuous casting dataset of a single material is preset for the first model training. Since the simulation data is massive, easily accessible, accurate and comprehensive, the prediction accuracy of the ingot strain and stress prediction model is further ensured. At the same time, the experimental continuous casting dataset of the target material is used for the second model training. There is no need for scene-by-scene modeling, which reduces the workload. Based on the transfer learning method, the ingot strain and stress prediction model can quickly adapt to new data. In the case of limited experimental data of the target material, the existing experimental data can be effectively utilized, which reduces the dependence on a large amount of new data, improves the generalization ability of the model, and further improves the prediction accuracy of the ingot strain and stress prediction model.

[0093] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are listed below. BRIEF DESCRIPTION OF THE DRAWINGS

[0094] Various other advantages and benefits will become apparent to those of ordinary skill in the art by reading the detailed description of the preferred embodiments below. The accompanying drawings are only for the purpose of illustrating the preferred embodiments and are not to be considered as limiting the present application. Also, the same reference symbols are used throughout the accompanying drawings to represent the same components. In the accompanying drawings:

[0095] Figure 1 A flow chart of a method for predicting stress and strain of a casting blank during a continuous casting process provided by an embodiment of the present application is shown;

[0096] Figure 2 A training flow chart of a strain stress prediction model for a casting blank provided in an embodiment of the present application is shown;

[0097] Figure 3 A block diagram of a device for predicting stress and strain of a casting blank during a continuous casting process provided by an embodiment of the present application is shown;

[0098] Figure 4 A schematic diagram of the structure of a terminal provided in an embodiment of the present application is shown. DETAILED DESCRIPTION

[0099] The exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although the exemplary embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments set forth herein. On the contrary, these embodiments are provided to enable a more thorough understanding of the present disclosure and to fully convey the scope of the present disclosure to those skilled in the art.

[0100] At the same time, it should be understood that for the convenience of description, the sizes of the various parts shown in the drawings are not drawn according to the actual proportional relationship.

[0101] The following description of at least one exemplary embodiment is merely illustrative in nature and is in no way intended to limit the present application, its application, or uses.

[0102] Technologies, methods, and equipment known to ordinary technicians in the relevant art may not be discussed in detail, but where appropriate, the technologies, methods, and equipment should be considered as part of the specification.

[0103] It should be noted that like reference numerals and letters refer to similar items in the following figures, and therefore, once an item is defined in one figure, it need not be further discussed in subsequent figures.

[0104] The embodiments of the present application can be applied to computer systems / servers, which can operate with many other general or special computing system environments or configurations. Examples of well-known computing systems, environments and / or configurations suitable for use with computer systems / servers include, but are not limited to: personal computer systems, server computer systems, thin clients, thick clients, handheld or laptop devices, microprocessor-based systems, set-top boxes, programmable consumer electronics, network personal computers, small computer systems, large computer systems, and distributed cloud computing technology environments including any of the above systems, etc.

[0105] Computer systems / servers may be described in the general context of computer system executable instructions (such as program modules) executed by a computer system. In general, program modules may include routines, programs, object programs, components, logic, data structures, etc., which perform specific tasks or implement specific abstract data types. Computer systems / servers may be implemented in a distributed cloud computing environment where tasks are performed by remote processing devices linked through a communication network. In a distributed cloud computing environment, program modules may be located on local or remote computing system storage media including storage devices.

[0106] The present application embodiment provides a method for predicting the stress and strain of a casting during a continuous casting process. Figure 1 As shown, the method includes:

[0107] 101. Obtain continuous casting process parameters and casting machine parameters for target materials.

[0108] The target material can be any metal, a single steel material, or an alloy material; the continuous casting process parameters can include material composition, solid phase ratio, reduction, drawing speed, etc.; the casting machine parameters can include injection pressure, casting pressure, output billet size, etc. In the embodiment of the present application, the current execution end can be a stress-strain prediction module in a continuous casting production control system, which obtains the continuous casting process parameters and casting machine parameters of the target material currently being continuously cast.

[0109] It should be noted that the target material can be a new material with only a small amount of experimental data, or it can be an existing material with a large amount of data, which is not specifically limited in the embodiments of the present application.

[0110] 102. Based on the strain and stress prediction model of the ingot that has completed model training, the strain and stress distribution data of the ingot of the target material is predicted according to the continuous casting process parameters and casting machine parameters.

[0111] Among them, the ingot strain and stress prediction model is constructed based on a conditional generative adversarial network, and is obtained by performing two model trainings using a simulated continuous casting data set of a preset single material and an experimental continuous casting data set of a target material. The preset single material is used to characterize a single type of steel with a large amount of simulation data, and the target material is used to characterize a material with only a small amount of experimental data, which can be any material; and the loss function used in the model training is corrected by the numerical deviation between the predicted ingot strain and stress distribution data and the simulated ingot strain and stress distribution data, as well as the curve morphology deviation. In an embodiment of the present application, a simulated continuous casting data set of a single steel grade with a large amount of simulation data can be first used to preliminarily train an intermediate ingot strain and stress prediction model, and the intermediate ingot strain and stress prediction model already has the ability to predict the strain and stress distribution data of the ingot of the single steel grade. Then, an experimental continuous casting data set of a target material with only a small amount of experimental data is used, based on a transfer learning method, to further train the final ingot strain and stress prediction model, so that the model also has the ability to predict the strain and stress distribution data of the ingot of the target material, and can achieve the technical effect of reducing the workload without the need for scene-based modeling. At the same time, the ingot strain and stress prediction model can quickly adapt to new data, and when the experimental data of the target material is limited, the existing experimental data can be effectively used, which reduces the dependence on a large amount of new data, improves the generalization ability of the model, and further improves the prediction accuracy of the ingot strain and stress prediction model. In addition, during the model training process, adversarial training is performed using the real data obtained from simulation and the generated predicted data. Since the essence of adversarial training is to generate predicted strain and stress distribution data of the ingot based on the generator, and then to judge the authenticity of the predicted strain and stress distribution data of the ingot based on the discriminator, by training the generator and the discriminator at the same time, the generator continuously improves the realism of the predicted data, and the discriminator continuously improves its ability to distinguish true and false data, thereby improving the accuracy and stability of the ingot strain and stress prediction model, thereby improving the prediction accuracy of ingot stress and strain during the continuous casting process.

[0112] Compared with the prior art, the embodiment of the present application constructs a strain and stress prediction model for the ingot based on a conditional generative adversarial network, and uses the real data obtained by simulation and the generated prediction data for adversarial training. Since the adversarial training is essentially based on the generator to generate predicted ingot strain and stress distribution data, and then based on the discriminator to discriminate the authenticity of the predicted ingot strain and stress distribution data, by training the generator and the discriminator at the same time, the generator continuously improves the fidelity of the prediction data, and the discriminator continuously improves the ability to distinguish true and false data, thereby improving the accuracy and stability of the ingot strain and stress prediction model, thereby improving the prediction accuracy of the ingot stress and strain in the continuous casting process; further, using A simulated continuous casting dataset of a single material is preset for the first model training. Since the simulation data is massive, easily accessible, accurate and comprehensive, the prediction accuracy of the ingot strain and stress prediction model is further ensured. At the same time, the experimental continuous casting dataset of the target material is used for the second model training. There is no need for scene-by-scene modeling, which reduces the workload. Based on the transfer learning method, the ingot strain and stress prediction model can quickly adapt to new data. In the case of limited experimental data of the target material, the existing experimental data can be effectively utilized, which reduces the dependence on a large amount of new data, improves the generalization ability of the model, and further improves the prediction accuracy of the ingot strain and stress prediction model.

[0113] In an embodiment of the present application, in order to further define and illustrate, Figure 2 As shown, before step 102 of the embodiment predicts the strain and stress distribution data of the target material based on the strain and stress prediction model of the cast billet for which model training has been completed and according to the continuous casting process parameters and the casting machine parameters, the embodiment method further includes:

[0114] 201. An initial generator is constructed based on a long short-term memory network, an initial discriminator is constructed based on a convolutional neural network, and the initial generator and the initial discriminator are combined to construct an initial ingot strain stress prediction model.

[0115] Among them, the generator is used to generate predicted ingot strain and stress distribution data according to the continuous casting process parameters and the casting machine parameters; the discriminator is used to discriminate the authenticity of the predicted ingot strain and stress distribution data according to the simulated ingot strain and stress distribution data in the simulated continuous casting data set; it should be noted that, based on the ingot strain and stress prediction model that has completed model training, the ingot strain and stress distribution data of the target material are predicted according to the continuous casting process parameters and the casting machine parameters. Specifically, based on the generator in the ingot strain and stress prediction model that has completed model training, the ingot strain and stress distribution data of the target material are predicted according to the continuous casting process parameters and the casting machine parameters.

[0116] In the embodiment of the present application, the basic unit of the long short-term memory network in the generator includes an input gate, an output gate, and a forget gate. The calculation inside the unit is shown in the following formula:

[0117]

[0118]

[0119] in, represents the input vector at time t, represents the forget gate at time t, represents the input gate at time t, represents the output gate at time t, represents the temporary state of the memory unit at time t, represents the output value of the hidden layer at time t, represents the state of the memory unit at time t, , , , Both represent weight matrices, , , , Both represent the offset. represents the sigmoid activation function, Indicates element-wise multiplication. The activation function of the discriminator can use the LeakyRrlu activation function, which can be expressed as the following formula:

[0120]

[0121] in, Represents a number close to 0, such as 0.02, Represents the input feature vector.

[0122] Finally, the initial generator and the initial discriminator are combined to construct an initial ingot strain and stress prediction model. When using the prediction model, the generator generates predicted ingot strain and stress distribution data based on the continuous casting process parameters and the casting machine parameters, and then the discriminator determines the authenticity of the predicted ingot strain and stress distribution data. The real data obtained by simulation and the generated prediction data are used for adversarial training, so that the generator continuously improves the realism of the prediction data, and the discriminator continuously improves the ability to distinguish true and false data, thereby improving the accuracy and stability of the ingot strain and stress prediction model, thereby improving the prediction accuracy of ingot stress and strain in the continuous casting process.

[0123] 202. Obtain a simulated continuous casting data set of a preset single material.

[0124] The simulated continuous casting data set includes the first continuous casting process parameters, the first casting machine parameters, and the first simulated ingot strain and stress distribution data for a preset single material. Specifically, firstly, the first casting machine parameters and the first continuous casting process parameters of a single steel grade are obtained; further, a geometric model of the ingot is established based on the output ingot size in the first casting machine parameters, which can be completed using drawing software such as CAD; further, the geometric model is imported into the finite element software, and based on the first continuous casting process parameters and the geometric model, the finite element software is used to perform simulation operations, thereby obtaining the simulated continuous casting data set of the single steel grade.

[0125] Correspondingly, step 202 of the embodiment specifically includes: obtaining the first casting machine parameters and the first continuous casting process parameters of the preset single material; establishing a geometric model of the ingot based on the first casting machine parameters; and performing simulation operations using finite element software based on the first continuous casting process parameters and the geometric model to obtain a simulated continuous casting data set of the preset single material.

[0126] 203. Based on the initial generator, according to the first continuous casting process parameters and the first casting machine parameters, first predicted cast billet strain stress distribution data of a preset single material is predicted.

[0127] The first predicted ingot strain and stress distribution data is used to characterize the ingot strain and stress distribution data predicted by the model, and is false distribution data.

[0128] 204. Based on the initial discriminator, according to the first continuous casting process parameters and the first casting machine parameters, the first simulated ingot strain and stress distribution data and the first predicted ingot strain and stress distribution data are distinguished to obtain a first authenticity discrimination result.

[0129] In the embodiment of the present application, false distribution data and real distribution data (simulated data) are distinguished based on an initial discriminator.

[0130] 205. Calculate a first loss function between the first authenticity discrimination result and the authenticity result.

[0131] Specifically, first, the first loss function between the first authenticity discrimination result and the authenticity result is calculated according to the following formula:

[0132]

[0133] in, Represents the noise distribution The expected value of Indicates the strain stress distribution data of the first simulation casting expected value.

[0134] Preferably, the accuracy of the model is further improved, and the first loss function is corrected based on the numerical deviation of the distribution data and the curve shape deviation. Specifically, the numerical deviation between the first predicted cast billet strain and stress distribution data and the first simulated cast billet strain and stress distribution data is calculated according to the following formula:

[0135]

[0136] in, Indicates the numerical deviation. Indicates the first predicted slab strain stress distribution data elements, The first simulation shows the strain stress distribution data of the first slab. elements, n represents the number of cross-sectional strain distribution points;

[0137] And according to the following formula, the curve shape deviation between the first predicted cast billet strain stress distribution data and the first simulated cast billet strain stress distribution data is calculated,

[0138]

[0139] in, Indicates the curve shape deviation, The first simulation shows the difference component d of the strain stress distribution data of the first simulation casting. The value of the dimension, Represents the first prediction of the difference component of the strain stress distribution data of the slab No. The value of the dimension;

[0140] Finally, for the first loss function , numerical deviation , curve shape deviation Perform weighted summation to obtain the modified first loss function, that is,

[0141]

[0142] in, represents the modified first loss function, Indicates the numerical deviation The weight parameter in the modified first loss function, Indicates the curve shape deviation Weight parameter in the modified first loss function.

[0143] It should be noted that in the embodiment of the present application, by introducing a numerical deviation based on Euclidean distance and a curve shape deviation based on cosine distance, the model loss function is corrected for deviation, thereby further improving the accuracy of the model and ensuring the consistency of the generated data in terms of value and shape.

[0144] 206. Update the model parameters of the initial generator and the initial discriminator based on the first loss function to obtain an intermediate generator and an intermediate discriminator.

[0145] Among them, the intermediate generator and the intermediate discriminator are used to characterize the prediction model obtained by preliminary model training using a simulated continuous casting data set of a preset single material. In the embodiment of the present application, the model parameters of the initial generator and the initial discriminator can be updated by solving the minimum objective function of the modified first loss function, or by setting the number of iterations, or by setting the prediction accuracy, which is not specifically limited in the embodiment of the present application.

[0146] 207. Obtain experimental continuous casting data set of target material.

[0147] The experimental continuous casting data set includes the second continuous casting process parameters, the second casting machine parameters, and the experimental ingot strain and stress distribution data for the target material. In the embodiment of the present application, the target material may be the same as or different from the above-mentioned preset single material. The data included in the experimental continuous casting data set are all real data obtained from the experiment.

[0148] 208. Based on the intermediate generator, according to the second continuous casting process parameters and the second casting machine parameters, the second predicted ingot strain stress distribution data of the target material is predicted.

[0149] The second predicted ingot strain and stress distribution data are used to characterize the ingot strain and stress distribution data predicted by the model, and are false distribution data.

[0150] 209. Based on the intermediate discriminator, according to the second continuous casting process parameters and the second casting machine parameters, the experimental ingot strain and stress distribution data and the second predicted ingot strain and stress distribution data are distinguished to obtain a second authenticity discrimination result.

[0151] In the embodiment of the present application, false distribution data and real distribution data (experimental data) are distinguished based on an intermediate discriminator.

[0152] 210. Calculate a second loss function between the second authenticity discrimination result and the authenticity result.

[0153] The specific implementation steps in the embodiment of the present application are the same as step 205 of the embodiment and will not be repeated here.

[0154] 211. Update the model parameters of the intermediate generator and the intermediate discriminator based on the second loss function to obtain the generator and the discriminator that have completed model training, so as to predict the strain and stress distribution data of the ingot based on the continuous casting process parameters and the casting machine parameters of the generator that has completed model training.

[0155] The specific implementation steps in the embodiment of the present application are the same as step 206 of the embodiment, and will not be repeated here.

[0156] In one embodiment of the present application, for further definition and explanation, before steps 203 and 208 of the embodiment, the embodiment method also includes: performing Z-Score standardization processing on the data contained in the simulation continuous casting data set so that the mean of each data is zero and the variance is 1, and performing prediction processing based on the initial generator and the standardized simulation continuous casting data set; and performing Z-Score standardization processing on the data contained in the experimental continuous casting data set so that the mean of each data is zero and the variance is 1, and performing prediction processing based on the intermediate generator and the standardized experimental continuous casting data set.

[0157] It should be noted that, in the continuous casting process parameters, the solid phase ratio represents the liquid content of molten steel, which can take values ​​of 0.6, 0.7, 0.8, 0.9, 1.0, etc., without units; the reduction amount represents the reduction in the height of the billet before and after continuous casting, which can take values ​​of 0.5, 1.0, 1.5, 2.0, 2.5, 3.0, etc., in units of %; the pulling speed represents the speed at which the billet is pulled out of the crystallizer, which can take values ​​of 0.41, 0.42, 0.43, 0.44, 0.45, etc., in units of m / min. Since the numerical order of magnitude of the above data differs greatly and the units are not uniform, in order to ensure the comparability between the data, in the embodiment of the present application, preferably, the data can be subjected to Z-Score standardization before being input into the model. Specifically, the experimental continuous casting data in each parameter is subtracted from the mean of the parameter, and then divided by the standard deviation of the parameter to obtain the standardized experimental continuous casting data. So that the effect of model training based on the standardized experimental continuous casting data set is more effective.

[0158] The present application provides a method for predicting stress and strain of a billet during a continuous casting process. First, continuous casting process parameters and casting machine parameters for a target material are obtained. Furthermore, based on a billet strain and stress prediction model for which model training has been completed, according to the continuous casting process parameters and the casting machine parameters, the billet strain and stress distribution data of the target material are predicted. The billet strain and stress prediction model is constructed based on a conditional generative adversarial network, and is obtained by performing two model trainings in sequence using a simulated continuous casting data set of a preset single material and an experimental continuous casting data set of the target material. The loss function used in the model training is obtained by correcting the numerical deviation and curve morphology deviation between the predicted billet strain and stress distribution data and the simulated billet strain and stress distribution data. Compared with the prior art, the embodiment of the present application constructs a strain and stress prediction model for the ingot based on a conditional generative adversarial network, and uses the real data obtained by simulation and the generated prediction data for adversarial training. Since the adversarial training is essentially based on the generator to generate predicted ingot strain and stress distribution data, and then based on the discriminator to discriminate the authenticity of the predicted ingot strain and stress distribution data, by training the generator and the discriminator at the same time, the generator continuously improves the fidelity of the prediction data, and the discriminator continuously improves the ability to distinguish true and false data, thereby improving the accuracy and stability of the ingot strain and stress prediction model, thereby improving the prediction accuracy of the ingot stress and strain in the continuous casting process; further, using A simulated continuous casting dataset of a single material is preset for the first model training. Since the simulation data is massive, easily accessible, accurate and comprehensive, the prediction accuracy of the ingot strain and stress prediction model is further ensured. At the same time, the experimental continuous casting dataset of the target material is used for the second model training. There is no need for scene-by-scene modeling, which reduces the workload. Based on the transfer learning method, the ingot strain and stress prediction model can quickly adapt to new data. In the case of limited experimental data of the target material, the existing experimental data can be effectively utilized, which reduces the dependence on a large amount of new data, improves the generalization ability of the model, and further improves the prediction accuracy of the ingot strain and stress prediction model.

[0159] Furthermore, as a response to the above Figure 1 The implementation of the method shown in the embodiment of the present application provides a prediction device for the stress and strain of the casting during the continuous casting process, such as Figure 3 As shown, the device comprises:

[0160] Parameter acquisition module 31, strain stress prediction module 32.

[0161] A parameter acquisition module is used to obtain continuous casting process parameters and casting machine parameters for target materials;

[0162] The strain and stress prediction module is used to predict the strain and stress distribution data of the target material based on the strain and stress prediction model of the cast billet that has completed model training, according to the continuous casting process parameters and the casting machine parameters. The strain and stress prediction model of the cast billet is constructed based on a conditional generative adversarial network, and is obtained by performing two model trainings in turn using a simulation continuous casting data set of a preset single material and an experimental continuous casting data set of the target material.

[0163] In a specific application scenario, the cast slab strain stress prediction model includes a generator and a discriminator;

[0164] The generator is used to generate predicted strain stress distribution data of the ingot according to the continuous casting process parameters and the casting machine parameters;

[0165] The discriminator is used to discriminate the authenticity of the predicted slab strain and stress distribution data according to the simulated slab strain and stress distribution data in the simulated continuous casting data set;

[0166] The strain stress prediction module is used to:

[0167] Based on the generator in the slab strain and stress prediction model for which model training has been completed, the slab strain and stress distribution data of the target material is predicted according to the continuous casting process parameters and the casting machine parameters.

[0168] In a specific application scenario, before the strain stress prediction module, the device further includes a model training module, including:

[0169] A model building unit, used to build an initial generator based on a long short-term memory network, build an initial discriminator based on a convolutional neural network, and combine the initial generator and the initial discriminator to build an initial casting billet strain stress prediction model;

[0170] A simulated continuous casting data set acquisition unit, used to acquire a simulated continuous casting data set of a preset single material, wherein the simulated continuous casting data set includes a first continuous casting process parameter, a first casting machine parameter, and a first simulated cast billet strain stress distribution data for the preset single material;

[0171] A first prediction unit is used to predict and obtain first predicted cast slab strain stress distribution data of the preset single material based on the initial generator, according to the first continuous casting process parameters and the first casting machine parameters;

[0172] A first discriminating unit is configured to discriminate the first simulated ingot strain and stress distribution data and the first predicted ingot strain and stress distribution data based on the initial discriminator and the first continuous casting process parameter and the first casting machine parameter to obtain a first authenticity discrimination result;

[0173] A first loss function calculation unit, used to calculate a first loss function between the first authenticity discrimination result and the authenticity result;

[0174] A first parameter updating unit, used for updating the model parameters of the initial generator and the initial discriminator based on the first loss function to obtain an intermediate generator and an intermediate discriminator;

[0175] An experimental continuous casting data set acquisition unit, used to acquire an experimental continuous casting data set of the target material, wherein the experimental continuous casting data set includes a second continuous casting process parameter and a second casting machine parameter for the target material, and experimental cast billet strain stress distribution data;

[0176] A second prediction unit is used to predict and obtain second predicted cast slab strain stress distribution data of the target material based on the intermediate generator, according to the second continuous casting process parameters and the second casting machine parameters;

[0177] A second discrimination unit is used to distinguish the experimental ingot strain and stress distribution data and the second predicted ingot strain and stress distribution data based on the intermediate discriminator and according to the second continuous casting process parameters and the second casting machine parameters, so as to obtain a second authenticity discrimination result;

[0178] A second loss function calculation unit, used to calculate a second loss function between the second authenticity discrimination result and the authenticity result;

[0179] The second parameter updating unit is used to update the model parameters of the intermediate generator and the intermediate discriminator based on the second loss function to obtain the generator and the discriminator that have completed the model training, so as to predict the strain and stress distribution data of the ingot based on the continuous casting process parameters and the casting machine parameters of the generator that has completed the model training.

[0180] In a specific application scenario, before the first parameter updating unit, the model training module further includes a first loss function correction unit, which is used to:

[0181] The first numerical deviation between the first predicted cast billet strain stress distribution data and the first simulated cast billet strain stress distribution data is calculated according to the following formula:

[0182]

[0183] in, Indicates the numerical deviation. Indicates the first predicted slab strain stress distribution data elements, The first simulation shows the strain stress distribution data of the first slab. elements, n represents the number of cross-sectional strain distribution points;

[0184] The first curve shape deviation between the first predicted cast billet strain stress distribution data and the first simulated cast billet strain stress distribution data is calculated according to the following formula:

[0185]

[0186] in, Indicates the curve shape deviation, The first simulation shows the difference component d of the strain stress distribution data of the first simulation casting. The value of the dimension, Represents the first predicted slab strain stress distribution data difference component No. The value of the dimension;

[0187] Correcting the first loss function according to the first numerical deviation and the first curve shape deviation to obtain a corrected first loss function, and updating the model parameters of the initial generator and the initial discriminator based on the corrected first loss function;

[0188] And before the second parameter updating unit, the model training module further includes a second loss function correction unit, which is used to:

[0189] According to the above formula, a second numerical deviation between the second predicted cast billet strain stress distribution data and the experimental cast billet strain stress distribution data is calculated;

[0190] According to the above formula, a second curve shape deviation between the second predicted cast billet strain stress distribution data and the experimental cast billet strain stress distribution data is calculated;

[0191] The second loss function is corrected according to the second numerical deviation and the second curve morphology deviation to obtain a corrected second loss function, so as to update the model parameters of the intermediate generator and the intermediate discriminator based on the corrected second loss function.

[0192] In a specific application scenario, before the first prediction unit, the strain stress prediction module further includes:

[0193] A first standardization processing unit is used to perform Z-Score standardization processing on the data contained in the simulated continuous casting data set so that the mean of each data is zero and the variance is 1, so as to perform prediction processing based on the simulated continuous casting data set after the standardization processing based on the initial generator;

[0194] Before the second prediction unit, the strain stress prediction module further includes:

[0195] The second standardization processing unit is used to perform Z-Score standardization on the data contained in the experimental continuous casting data set so that the mean of each data is zero and the variance is 1, so as to perform prediction processing based on the standardized experimental continuous casting data set based on the intermediate generator.

[0196] In a specific application scenario, the simulation continuous casting data set acquisition unit is used to:

[0197] Acquire a first casting machine parameter and a first continuous casting process parameter of the preset single material;

[0198] Establishing a geometric model of the ingot based on the first casting machine parameters;

[0199] Based on the first continuous casting process parameters and the geometric model, a simulation operation is performed using finite element software to obtain a simulation continuous casting data set of the preset single material.

[0200] In a specific application scenario, the continuous casting process parameters include material composition, solid phase ratio, reduction amount, and casting speed.

[0201] The present application provides a device for predicting the stress and strain of a cast billet during a continuous casting process. First, the continuous casting process parameters and casting machine parameters for a target material are obtained. Furthermore, based on a cast billet strain and stress prediction model for which model training has been completed, the cast billet strain and stress distribution data of the target material is predicted according to the continuous casting process parameters and the casting machine parameters. The cast billet strain and stress prediction model is constructed based on a conditional generative adversarial network, and is obtained by performing two model trainings in sequence using a simulated continuous casting data set of a preset single material and an experimental continuous casting data set of the target material. The loss function used in the model training is obtained by correcting the numerical deviation and curve morphology deviation between the predicted cast billet strain and stress distribution data and the simulated cast billet strain and stress distribution data. Compared with the prior art, the embodiment of the present application constructs a strain and stress prediction model for the ingot based on a conditional generative adversarial network, and uses the real data obtained by simulation and the generated prediction data for adversarial training. Since the adversarial training is essentially based on the generator to generate predicted ingot strain and stress distribution data, and then based on the discriminator to discriminate the authenticity of the predicted ingot strain and stress distribution data, by training the generator and the discriminator at the same time, the generator continuously improves the fidelity of the prediction data, and the discriminator continuously improves the ability to distinguish true and false data, thereby improving the accuracy and stability of the ingot strain and stress prediction model, thereby improving the prediction accuracy of the ingot stress and strain in the continuous casting process; further, using A simulated continuous casting dataset of a single material is preset for the first model training. Since the simulation data is massive, easily accessible, accurate and comprehensive, the prediction accuracy of the ingot strain and stress prediction model is further ensured. At the same time, the experimental continuous casting dataset of the target material is used for the second model training. There is no need for scene-by-scene modeling, which reduces the workload. Based on the transfer learning method, the ingot strain and stress prediction model can quickly adapt to new data. In the case of limited experimental data of the target material, the existing experimental data can be effectively utilized, which reduces the dependence on a large amount of new data, improves the generalization ability of the model, and further improves the prediction accuracy of the ingot strain and stress prediction model.

[0202] According to one embodiment of the present application, a storage medium is provided, wherein the storage medium stores at least one executable instruction, and the computer executable instruction can execute the method for predicting stress and strain of the ingot during the continuous casting process in any of the above method embodiments.

[0203] Based on this understanding, the technical solution of the present application can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (which can be a CD-ROM, a USB flash drive, a mobile hard disk, etc.), and includes a number of instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each implementation scenario of the present application.

[0204] Figure 4A schematic diagram of the structure of a terminal provided according to an embodiment of the present application is shown, and the specific embodiment of the present application does not limit the specific implementation of the terminal.

[0205] like Figure 4 As shown, the terminal may include: a processor (processor) 402 , a communication interface (Communications Interface) 404 , a memory (memory) 406 , and a communication bus 408 .

[0206] The processor 402 , the communication interface 404 , and the memory 406 communicate with each other via a communication bus 408 .

[0207] The communication interface 404 is used to communicate with other devices such as clients or other servers.

[0208] The processor 402 is used to execute the program 410, and specifically can execute the relevant steps in the above-mentioned embodiment of the method for predicting the stress and strain of the casting during the continuous casting process.

[0209] Specifically, the program 410 may include program codes, which include computer operation instructions.

[0210] The processor 402 may be a central processing unit (CPU), or an application-specific integrated circuit (ASIC), or one or more integrated circuits configured to implement the embodiments of the present application. The one or more processors included in the computer device may be processors of the same type, such as one or more CPUs; or may be processors of different types, such as one or more CPUs and one or more ASICs.

[0211] The memory 406 is used to store the program 410. The memory 406 may include a high-speed RAM memory, and may also include a non-volatile memory (non-volatile memory), such as at least one disk memory.

[0212] The program 410 may be specifically configured to enable the processor 402 to perform the following operations:

[0213] Obtain continuous casting process parameters and casting machine parameters for target materials;

[0214] Based on the ingot strain and stress prediction model that has completed model training, the ingot strain and stress distribution data of the target material is predicted according to the continuous casting process parameters and the casting machine parameters. The ingot strain and stress prediction model is constructed based on a conditional generative adversarial network, and is trained twice using a preset single material simulation continuous casting data set and an experimental continuous casting data set of the target material in sequence. The loss function used in the model training is corrected by the numerical deviation and curve morphology deviation between the predicted ingot strain and stress distribution data and the simulated ingot strain and stress distribution data.

[0215] The storage medium may also include an operating system and a network communication module. The operating system is a program that manages the hardware and software resources of the physical device of the above-mentioned method for predicting the stress and strain of the cast slab during the continuous casting process, and supports the operation of the information processing program and other software and / or programs. The network communication module is used to realize the communication between the components inside the storage medium, and the communication with other hardware and software in the information processing physical device.

[0216] Each embodiment in this specification is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the embodiments can be referred to each other. For the system embodiment, since it basically corresponds to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.

[0217] The method and system of the present application may be implemented in many ways. For example, the method and system of the present application may be implemented by software, hardware, firmware, or any combination of software, hardware, and firmware. The above order of steps for the method is for illustration only, and the steps of the method of the present application are not limited to the order specifically described above, unless otherwise specifically stated. In addition, in some embodiments, the present application may also be implemented as a program recorded in a recording medium, which includes machine-readable instructions for implementing the method according to the present application. Therefore, the present application also covers a recording medium storing a program for executing the method according to the present application.

[0218] Obviously, those skilled in the art should understand that the above modules or steps of the present application can be implemented by a general computing device, they can be concentrated on a single computing device, or distributed on a network composed of multiple computing devices, and optionally, they can be implemented by a program code executable by a computing device, so that they can be stored in a storage device and executed by the computing device, and in some cases, the steps shown or described can be executed in a different order from that herein, or they can be made into individual integrated circuit modules, or multiple modules or steps therein can be made into a single integrated circuit module for implementation. Thus, the present application is not limited to any specific combination of hardware and software.

[0219] The above description is only the preferred embodiment of the present application and is not intended to limit the present application. For those skilled in the art, the present application may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A method for predicting stress and strain of a casting during continuous casting, characterized in that: include: Obtain continuous casting process parameters and casting machine parameters for target materials; Based on the ingot strain and stress prediction model that has completed model training, the ingot strain and stress distribution data of the target material is predicted according to the continuous casting process parameters and the casting machine parameters. The ingot strain and stress prediction model is constructed based on a conditional generative adversarial network, and is trained twice using a preset single material simulation continuous casting data set and an experimental continuous casting data set of the target material in sequence. The loss function used in the model training is corrected by the numerical deviation and curve morphology deviation between the predicted ingot strain and stress distribution data and the simulated ingot strain and stress distribution data.

2. The method according to claim 1, characterized in that The cast slab strain stress prediction model includes a generator and a discriminator; The generator is used to generate predicted strain stress distribution data of the ingot according to the continuous casting process parameters and the casting machine parameters; The discriminator is used to discriminate the authenticity of the predicted slab strain and stress distribution data according to the simulated slab strain and stress distribution data in the simulated continuous casting data set; The slab strain stress prediction model based on the completed model training predicts the slab strain stress distribution data of the target material according to the continuous casting process parameters and the casting machine parameters, including: Based on the generator in the slab strain and stress prediction model for which model training has been completed, the slab strain and stress distribution data of the target material is predicted according to the continuous casting process parameters and the casting machine parameters.

3. The method according to claim 2, characterized in that Before the strain and stress prediction model of the cast billet based on the completed model training predicts the strain and stress distribution data of the cast billet of the target material according to the continuous casting process parameters and the casting machine parameters, the method further comprises: Constructing an initial generator based on a long short-term memory network, constructing an initial discriminator based on a convolutional neural network, and combining the initial generator and the initial discriminator to construct an initial casting billet strain stress prediction model; Acquire a simulated continuous casting data set of a preset single material, wherein the simulated continuous casting data set includes a first continuous casting process parameter, a first casting machine parameter, and a first simulated cast billet strain stress distribution data for the preset single material; Based on the initial generator, according to the first continuous casting process parameters and the first casting machine parameters, predict and obtain first predicted cast billet strain stress distribution data of the preset single material; Based on the initial discriminator, according to the first continuous casting process parameters and the first casting machine parameters, the first simulated ingot strain and stress distribution data and the first predicted ingot strain and stress distribution data are distinguished to obtain a first authenticity discrimination result; Calculating a first loss function between the first authenticity discrimination result and the authenticity result; Update the model parameters of the initial generator and the initial discriminator based on the first loss function to obtain an intermediate generator and an intermediate discriminator; Acquire an experimental continuous casting data set of the target material, wherein the experimental continuous casting data set includes a second continuous casting process parameter, a second casting machine parameter, and experimental cast billet strain stress distribution data for the target material; Based on the intermediate generator, according to the second continuous casting process parameters and the second casting machine parameters, predict and obtain second predicted cast slab strain stress distribution data of the target material; Based on the intermediate discriminator, according to the second continuous casting process parameters and the second casting machine parameters, the experimental ingot strain and stress distribution data and the second predicted ingot strain and stress distribution data are distinguished to obtain a second authenticity discrimination result; Calculating a second loss function between the second authenticity discrimination result and the authenticity result; Based on the second loss function, the model parameters of the intermediate generator and the intermediate discriminator are updated to obtain the generator and the discriminator that have completed model training, so as to predict the strain and stress distribution data of the ingot based on the continuous casting process parameters and the casting machine parameters based on the generator that has completed model training.

4. The method according to claim 3, characterized in that Before updating the model parameters of the initial generator and the initial discriminator based on the first loss function to obtain an intermediate generator and an intermediate discriminator, the method further includes: The first numerical deviation between the first predicted cast billet strain stress distribution data and the first simulated cast billet strain stress distribution data is calculated according to the following formula: in, Indicates the numerical deviation. Indicates the first predicted slab strain stress distribution data elements, The first simulation shows the strain stress distribution data of the first slab. elements, n represents the number of cross-sectional strain distribution points; The first curve shape deviation between the first predicted cast billet strain stress distribution data and the first simulated cast billet strain stress distribution data is calculated according to the following formula: in, Indicates the curve shape deviation, The first simulation shows the difference component d of the strain stress distribution data of the first casting billet. The value of the dimension, Represents the first prediction of the difference component of the strain stress distribution data of the slab No. The value of the dimension; Correcting the first loss function according to the first numerical deviation and the first curve shape deviation to obtain a corrected first loss function, and updating the model parameters of the initial generator and the initial discriminator based on the corrected first loss function; And before updating the model parameters of the intermediate generator and the intermediate discriminator based on the second loss function to obtain the generator and the discriminator that have completed model training, the method also includes: According to the above formula, a second numerical deviation between the second predicted cast billet strain stress distribution data and the experimental cast billet strain stress distribution data is calculated; According to the above formula, a second curve shape deviation between the second predicted cast billet strain stress distribution data and the experimental cast billet strain stress distribution data is calculated; The second loss function is corrected according to the second numerical deviation and the second curve morphology deviation to obtain a corrected second loss function, so as to update the model parameters of the intermediate generator and the intermediate discriminator based on the corrected second loss function.

5. The method according to claim 3, characterized in that: Before predicting and obtaining the first predicted cast billet strain stress distribution data of the preset single material based on the initial generator and according to the first continuous casting process parameters and the first casting machine parameters, the method further includes: Performing Z-Score standardization on the data contained in the simulated continuous casting data set so that the mean of each data is zero and the variance is 1, so as to perform prediction processing based on the standardized simulated continuous casting data set based on the initial generator; Before predicting and obtaining the second predicted cast slab strain stress distribution data of the target material based on the intermediate generator and according to the second continuous casting process parameters and the second casting machine parameters, the method further includes: The data contained in the experimental continuous casting data set is subjected to Z-Score standardization processing so that the mean of each data is zero and the variance is 1, so as to perform prediction processing based on the standardized experimental continuous casting data set based on the intermediate generator.

6. The method according to claim 3, characterized in that: The step of obtaining a simulated continuous casting data set of a preset single material includes: Obtaining first casting machine parameters and first continuous casting process parameters of the preset single material; Establishing a geometric model of the ingot based on the first casting machine parameters; Based on the first continuous casting process parameters and the geometric model, a simulation operation is performed using finite element software to obtain a simulation continuous casting data set of the preset single material.

7. The method according to any one of claims 1 to 6, characterized in that: The continuous casting process parameters include material composition, solid phase ratio, reduction amount, and casting speed.

8. A device for predicting stress and strain of a slab during continuous casting, characterized in that: include: A parameter acquisition module is used to obtain continuous casting process parameters and casting machine parameters for target materials; A strain and stress prediction module is used to predict the strain and stress distribution data of the target material based on the strain and stress prediction model of the ingot that has completed model training, according to the continuous casting process parameters and the casting machine parameters. The strain and stress prediction model of the ingot is constructed based on a conditional generative adversarial network, and is obtained by performing two model trainings in sequence using a simulated continuous casting data set of a preset single material and an experimental continuous casting data set of the target material. The loss function used in the model training is corrected by the numerical deviation and curve morphology deviation between the predicted strain and stress distribution data of the ingot and the simulated strain and stress distribution data of the ingot.

9. A storage medium, wherein at least one executable instruction is stored in the storage medium, characterized in that: The executable instructions enable the processor to execute operations corresponding to the method for predicting stress and strain of a cast billet during a continuous casting process according to any one of claims 1-7.

10. A terminal, comprising: A processor, a memory, a communication interface and a communication bus, wherein the processor, the memory and the communication interface communicate with each other via the communication bus; The memory is used to store at least one executable instruction, wherein the executable instruction enables the processor to execute operations corresponding to the method for predicting stress and strain of a cast billet during a continuous casting process as described in any one of claims 1-7.

Images