Die temperature prediction method, device and equipment for integrated thin-wall part die-casting process and medium

Abstract

The invention discloses an integrated thin-wall part die-casting process die temperature prediction method, device and equipment and a medium, and relates to the technical field of casting manufacturing. The method comprises the steps that thermal parameters of a die in the die-casting process are obtained; performing numerical simulation processing on the thermal parameters to obtain an output result; the output result comprises an initial temperature field image; determining a noise vector according to the output result, and performing combination processing on the noise vector and the thermal parameter to obtain a combination result; inputting a combination result into a trained neural network model for prediction processing to obtain a target temperature field image; the target temperature field image comprises a plurality of target temperature values of the mold; the neural network model is obtained by performing adversarial training according to the initial generative adversarial network and a thermal parameter sample of the die in the die-casting process. According to the method, the multiple target temperature values of the mold can be accurately determined, the mold temperature prediction accuracy is improved, mold temperature control is achieved, and product consistency and stability are kept.

Description

Technical Field

[0001] The present application relates to the technical field of casting manufacturing, and in particular to a method, device, equipment and medium for predicting mold temperature in a die-casting process of an integrated thin-walled part. Background Art

[0002] With the rapid development of product manufacturing, one-piece thin-walled parts are widely used in aerospace, automotive industry, electronic equipment and other fields because of their lightweight, high strength and good surface quality. Among them, one-piece thin-walled parts refer to die-casting parts composed of a single body and with a wall thickness in the range of 2mm-7mm. As one of the important methods for one-piece thin-walled parts, the die-casting process has the advantages of high production efficiency, high dimensional accuracy and good surface quality. It quickly injects molten metal into the mold cavity under high pressure and cools and solidifies it to form a product of the desired shape. In the die-casting process, the temperature of the mold directly affects the quality and performance of the die-casting. The appropriate mold temperature can ensure a good filling effect, reduce thermal stress and shrinkage defects, and help improve the surface finish and mechanical strength. However, inappropriate mold temperature may lead to a series of problems, such as poor filling, cold shut, pores, cracks and other defects, as well as increased scrap rate and reduced production efficiency. Therefore, in order to ensure product quality, improve production efficiency and extend mold life, how to predict the mold temperature is crucial to the die-casting process.

[0003] At present, the relevant technology uses machine learning algorithms to predict mold temperature, and collects a large amount of historical data to train neural networks to predict the change trend of mold temperature. However, this solution requires a large amount of data to train the model. When the data quality is low, the prediction accuracy of mold temperature is low. Summary of the invention

[0004] The purpose of the present application is to provide a method, device, equipment and medium for predicting the mold temperature of an integrated thin-walled part die-casting process.

[0005] To achieve the above objectives, this application provides the following solutions: In a first aspect, the present application provides a method for predicting mold temperature in a die-casting process of an integrated thin-walled part, comprising: Obtain the thermal parameters of the mold during the die casting process; Performing numerical simulation processing on the thermal parameters to obtain output results; the output results include an initial temperature field image; Determine a noise vector according to the output result, and combine the noise vector and the thermal parameter to obtain a combination result; The combined result is input into a trained neural network model for prediction processing to obtain a target temperature field image; the target temperature field image includes multiple target temperature values ​​of the mold; the neural network model is obtained by adversarial training based on an initial generative adversarial network and thermal parameter samples of the mold during the die-casting process.

[0006] Optionally, the initial generative adversarial network includes a generator to be trained and a discriminator to be trained connected to the generator to be trained; the neural network model is constructed in the following manner: Acquiring thermal parameter samples and noise vector samples, and performing numerical simulation processing on the thermal parameter samples to obtain simulation results; Dividing the simulation results into a training set and a test set; Using the training set and the noise vector sample to train the parameters of the discriminator to be trained and the generator to be trained, to obtain a generative adversarial network to be verified; Using the test set to verify the to-be-verified generative adversarial network, and determining the to-be-verified generative adversarial network that has been successfully verified as a generative adversarial network; The generator in the generative adversarial network is used as the neural network model.

[0007] Optionally, using the training set and the noise vector sample to train parameters in the discriminator to be trained and the generator to be trained to obtain a generative adversarial network to be verified includes: Selecting a portion of the data set from the training set; Inputting the noise vector sample and the thermal parameter sample into the generator to be trained in parallel to obtain a simulated temperature field image; Performing feature extraction processing on the partial data set to obtain a real temperature field image; Inputting the real temperature field image and the simulated temperature field image into a discriminator to be trained to obtain a prediction result; the prediction result is used to characterize the probability that the real temperature field image or the simulated temperature field image is evaluated as a real sample; A loss function is constructed according to the prediction result, and the discriminator to be trained and the generator to be trained are trained in sequence according to minimization of the loss function to obtain the generative adversarial network to be verified.

[0008] Optionally, the noise vector sample and the thermal parameter sample are inputted into a generator to be trained in parallel to obtain a simulated temperature field image, including: Performing encoding processing on the thermal parameter samples to obtain thermal parameter encoding results; Inserting the hot parameter encoding result into the noise vector sample to obtain a generated vector; The generated vector is input into the generator to be trained to obtain the simulated temperature field image.

[0009] Optionally, the prediction result includes a first result and a second result; the real temperature field image and the simulated temperature field image are input into a discriminator to be trained to obtain a prediction result, including: Inputting the real temperature field image and the simulated temperature field image into the convolution layer of the discriminator to be trained respectively for feature extraction, so as to obtain first feature information corresponding to the real temperature field image and second feature information corresponding to the simulated temperature field image; The thermal parameter encoding result is fused with the first feature information to obtain first fused information, and the first fused information is input into the fully connected layer in the discriminator to be trained for processing to obtain the first result, which is a first probability value of the real temperature field image being evaluated as a real sample; The thermal parameter encoding result and the second feature information are fused in parallel to obtain second fused information, and the second fused information is input into the fully connected layer in the discriminator to be trained for processing to obtain the second result, which is a second probability value of the simulated temperature field image being evaluated as a real sample.

[0010] Optionally, the loss function includes: a pixel loss component, a perceptual loss component and an adversarial loss component; The pixel loss component is used to characterize the loss between the real temperature field image and the simulated temperature field image; The perceptual loss component is used to characterize the difference between the first hidden layer features of the real temperature field image and the second hidden layer features of the simulated temperature field image; the first hidden layer features are the output features of the hidden layer of the real temperature field image in the discriminator to be trained, and the second hidden layer features are the output features of the hidden layer of the simulated temperature field image in the discriminator to be trained.

[0011] Optionally, the thermal parameters include at least one of the following: thermal conductivity, specific heat capacity, latent heat of fusion, density, solidification range, solid fraction, thermal expansion coefficient, radiation coefficient, heat conduction boundary conditions and cooling rate.

[0012] In a second aspect, the present application provides an integrated thin-walled part die-casting process mold temperature prediction device, comprising: An acquisition module, used to acquire thermal parameters of the mold during the die-casting process; A numerical simulation module, used for performing numerical simulation processing on the thermal parameters to obtain output results; the output results include an initial temperature field image; A combination module, used for determining a noise vector according to the output result, and combining the noise vector and the thermal parameter to obtain a combination result; A prediction module is used to input the combination result into a trained neural network model for prediction processing to obtain a target temperature field image; the target temperature field image includes multiple target temperature values ​​of the mold; the neural network model is obtained by adversarial training based on an initial generative adversarial network and thermal parameter samples of the mold during the die-casting process.

[0013] In a third aspect, the present application provides a computer device comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of any one of the above-described methods for predicting mold temperature of a one-piece thin-walled part die-casting process.

[0014] In a fourth aspect, the present application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of any one of the above-described methods for predicting mold temperature in a die-casting process for an integrated thin-walled part.

[0015] According to the specific embodiments provided in this application, this application discloses the following technical effects: The present application provides a method, device, equipment and medium for predicting the mold temperature in a die-casting process of an integrated thin-walled part. The method obtains the thermal parameters of the mold during the die-casting process, performs numerical simulation processing on the thermal parameters, and obtains an output result, which includes an initial temperature field image. A noise vector is determined based on the output result, and the noise vector and the thermal parameters are combined to obtain a combined result. The combined result is then input into a trained neural network model for prediction processing to obtain a target temperature field image. The target temperature field image includes multiple target temperature values ​​of the mold. The neural network model is obtained by adversarial training based on an initial generative adversarial network and thermal parameter samples of the mold during the die-casting process. Compared with the prior art, in this solution, after obtaining the thermal parameters of the mold during the die-casting process, the thermal parameters are numerically simulated to obtain the initial temperature field image, and the image is used as a data set for model input, providing good data guidance information for subsequent model processing, and combining the noise vector with the thermal parameters to more comprehensively determine the combination result, so that the determined combination result integrates multiple aspects of information; and because the neural network model is obtained in advance based on the initial generative adversarial network and thermal parameter sample adversarial training, its training process is similar to the application process, that is, the numerical simulation processing results are used as the data set for model training, which improves the quality of the model's input data set, and then after the neural network model is trained, the combination result can be directly predicted by the neural network model, and by combining physical drive and data drive, multiple target temperature values ​​of the mold can be accurately determined, thereby improving the prediction accuracy of the mold temperature, further realizing the control of the mold temperature, and maintaining the consistency and stability of the product. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. 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 paying creative work.

[0017] Figure 1 An application environment diagram of a method for predicting mold temperature in a die-casting process of an integrated thin-walled part provided in one embodiment of the present application; Figure 2 A schematic flow chart of a method for predicting mold temperature in a die-casting process for an integrated thin-walled part provided in one embodiment of the present application; Figure 3 A schematic diagram of a method flow for constructing a neural network model provided in one embodiment of the present application; Figure 4 A schematic diagram of a processing flow for constructing a proxy model through a training set and a test set provided in an embodiment of the present application; Figure 5 A schematic diagram of a processing flow for obtaining a prediction result of x, y and z through a generator to be trained and a discriminator to be trained provided in an embodiment of the present application; Figure 6 A schematic diagram of a processing flow of a generator to be trained provided in an embodiment of the present application; Figure 7 A schematic diagram of a processing flow of a discriminator to be trained provided in an embodiment of the present application; Figure 8 A schematic diagram of the functional modules of a mold temperature prediction device for a die-casting process of an integrated thin-walled part provided in an embodiment of the present application; Fig. 9 A schematic diagram of the structure of a computer device provided in one embodiment of the present application. DETAILED DESCRIPTION

[0018] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0019] In order to make the above-mentioned objects, features and advantages of the present application more obvious and easy to understand, the present application is further described in detail below with reference to the accompanying drawings and specific implementation methods.

[0020] The mold temperature prediction method of the die-casting process of an integrated thin-walled part provided in the embodiment of the present application can be applied to Figure 1In the application environment shown. Among them, the terminal 102 communicates with the server 104 through the network. The data storage system can store the data that the server 104 needs to process. The data storage system can be set up separately, or it can be integrated on the server 104, or it can be placed on the cloud or other servers. The terminal 102 can send the thermal parameters of the mold during the die-casting process to the server 104. After the server 104 receives the thermal parameters of the mold during the die-casting process, it performs numerical simulation processing on the thermal parameters to obtain the output results, and determines the combined results, and performs prediction processing through the trained neural network model to obtain the target temperature field image. The server 104 can feed back the obtained target temperature field image to the terminal 102 to visualize the target temperature field image through the terminal 102. In addition, in some embodiments, the mold temperature prediction method for the die-casting process of an integrated thin-walled part can also be implemented by the server 104 or the terminal 102 alone, such as the terminal 102 can directly perform simulation processing and neural network model prediction processing according to the thermal parameters of the mold in the die-casting process to obtain a target temperature field image, or the server 104 can obtain the thermal parameters in the die-casting process from the data storage system, and perform simulation processing and neural network model prediction processing according to the thermal parameters of the mold in the die-casting process to obtain a target temperature field image. The neural network model can be pre-stored in the server 104.

[0021] The terminal 102 may be, but is not limited to, various desktop computers, laptop computers, smart phones, tablet computers, IoT devices, and portable wearable devices. The IoT devices may be smart speakers, smart TVs, smart air conditioners, smart vehicle-mounted devices, etc. The portable wearable devices may be smart watches, smart bracelets, head-mounted devices, etc. The server 104 may be implemented as an independent server or a server cluster consisting of multiple servers, or may be a cloud server.

[0022] In an exemplary embodiment, Figure 2 As shown, a method for predicting mold temperature of a one-piece thin-walled part die-casting process is provided. The method is executed by a computer device, and specifically can be executed by a computer device such as a terminal or a server alone, or can be executed by a terminal and a server together. In the embodiment of the present application, the method is applied to Figure 1 The server 104 in the example is used as an example to illustrate the method, which includes the following steps S201 to S204. Among them: Step S201, obtaining thermal parameters of the mold during the die-casting process.

[0023] It is understandable that the right mold temperature is crucial to obtain high-quality die castings. For example, when die-casting aluminum alloy, the right mold temperature can ensure that the aluminum alloy melt flows smoothly in the mold cavity to fill the entire cavity. If the mold temperature is too low, the fluidity of the melt will deteriorate, which may cause defects such as cold shut and insufficient pouring in the die casting.

[0024] Optionally, the thermal parameters of the mold during the die-casting process can be obtained by real-time measurement, or imported from an external device, or obtained from a blockchain or database, or obtained by consulting materials. This embodiment does not impose any limitation on the method of obtaining the thermal parameters.

[0025] The thermal parameters include at least one of the following: thermal conductivity, specific heat capacity, fusion latent heat, density, solidification range, solid fraction, thermal expansion coefficient, radiation coefficient, heat conduction boundary conditions and cooling rate.

[0026] It should be noted that thermal conductivity, also known as thermal conductivity, refers to the amount of heat transferred through an area of ​​1m² in 1s when the temperature difference between the two surfaces of a 1m thick material is 1k (Kelvin) under stable heat transfer conditions. It reflects the ability of the material to conduct heat. Specific heat capacity refers to the amount of heat required to raise the unit temperature of a unit mass of a substance. The specific heat capacity of the mold material determines the ability of the mold to absorb heat. Mold materials with higher specific heat capacity can absorb more heat during the die casting process, thereby slowing down the rate of increase in the mold temperature. The latent heat of solidification refers to the amount of heat released by a unit mass of a substance during the solidification process. When a substance changes from liquid to solid, heat is released. Different metals have different densities. During the die casting process, the density affects the flow state of the molten metal in the mold cavity. The coefficient of thermal expansion refers to the relative rate of change of the length, area or volume of an object when the temperature rises by 1k. The emissivity is the ratio of the radiation ability of the surface of an object to the radiation ability of a black body at the same temperature. It reflects the ability of an object to transfer heat by thermal radiation. Heat conduction boundary conditions refer to the prescribed conditions of physical quantities such as temperature and heat flux density at the boundary of an object during heat conduction. For example, the surface temperature and heat flux density of the mold in contact with the cooling medium (such as water). The cooling rate is the speed at which the temperature of an object decreases, and the unit is K / s or ℃ / s. A faster cooling rate can refine the grain structure of the casting and improve the mechanical properties of the casting. The solid fraction refers to the mass fraction or volume fraction of the solidified solid part of the entire alloy during the solidification process.

[0027] Step S202, numerical simulation processing is performed on the thermal parameters to obtain output results; the output results include an initial temperature field image.

[0028] After obtaining the thermal parameters, a solver can be used to perform numerical simulation on the thermal parameters. By using the following energy equation, an output result can be obtained. The output result may include an initial temperature field image. The initial temperature field image may include the mold temperature corresponding to the mold at different times: ; Where T is the mold temperature, ρ is the density, C p is the specific heat, k is the thermal conductivity, L f is the latent heat of fusion, f s is the solid fraction and t is the time.

[0029] Step S203, determining a noise vector according to the output result, and combining the noise vector and the thermal parameter to obtain a combined result.

[0030] It should be noted that the above-mentioned noise vector refers to a random variable determined from the noise distribution, and the combination result refers to the result of fusing the noise vector and the thermal parameter.

[0031] Specifically, after obtaining the output result, the noise vector can be determined, which is generally sampled from a simple probability distribution (such as normal distribution or uniform distribution) and can be a random variable of Gaussian distribution. For example, in many cases, z is a vector randomly drawn from a standard normal distribution N(0,1), and its dimension can vary depending on the specific generation task and model design.

[0032] Optionally, in the process of combining the noise vector and the thermal parameters to obtain the combined result, the noise vector and the thermal parameters may be concatenated; the thermal parameters may be encoded first to obtain the encoded result, and then the encoded result may be inserted into the noise vector to obtain the combined result; or the noise vector and the thermal parameters may be fused to obtain the combined result.

[0033] In this embodiment, by determining the noise vector according to the output result and combining the noise vector and the thermal parameters, the combination result can integrate multiple information, thereby making the generated temperature field image meet the corresponding thermal parameters and improving the accuracy of the target temperature value.

[0034] Step S204, input the combined result into the trained neural network model for prediction processing to obtain a target temperature field image; the target temperature field image includes multiple target temperature values ​​of the mold; the neural network model is obtained by adversarial training based on the initial generative adversarial network and the thermal parameter samples of the mold during the die-casting process.

[0035] It should be noted that the above neural network model is a model obtained by pre-training the thermal parameter samples of the mold during the die-casting process, and can predict the target temperature field image. The target temperature field image is a temperature field image that satisfies the thermal parameters, which may include multiple target temperature values, wherein each target temperature value can be represented by different colors in the target temperature field image, each color corresponds to a target temperature value, and different target temperature values ​​corresponding to different colors can be pre-customized according to actual needs. The initial generative adversarial network is a generative adversarial network whose parameters are to be optimized, and the thermal parameter samples are samples used to train the initial generative adversarial network, which may include: thermal conductivity, specific heat capacity, latent heat of fusion, density, solidification range, solid fraction, thermal expansion coefficient, radiation coefficient, thermal conduction boundary conditions and cooling rate. The neural network model is a generator obtained after adversarial training of the initial generative adversarial network.

[0036] The neural network model may include multiple neural network layers, such as convolutional layers, and the multiple convolutional layers may be connected in sequence. Specifically, after obtaining the combination result, the combination result may be input into the trained neural network model, and the combination result may be subjected to feature extraction through multiple convolutional layers in sequence to obtain the target temperature field image.

[0037] In this embodiment, the above-mentioned initial generative adversarial network can be a GAN model, which includes: a generator (G) and a discriminator (D). In the process of training the GAN model, the training goal of the generator is to generate data that is as real as possible, while the discriminator needs to distinguish between real data and data generated by the generator. Through this adversarial training method, the generator is continuously optimized and can eventually generate new data with a distribution similar to that of the training data. The output of the discriminator is a value between 0 and 1, indicating the probability that the input data is real data.

[0038] Among them, the mold temperature prediction method for the die-casting process of integrated thin-walled parts provided in the embodiment of the present application can be applied to the mold temperature prediction scenario in the die-casting process of integrated thin-walled parts, and specifically can be applied to the mold temperature prediction scenario in the die-casting process of integrated large thin-walled parts. An integrated large thin-walled part refers to a die-casting part composed of a single main body, with a wall thickness in the range of 2mm-7mm and a cross-sectional area greater than 1.2 square meters.

[0039] The present application provides a method for predicting the mold temperature of an integrated thin-walled part in a die-casting process. The method obtains the thermal parameters of the mold during the die-casting process, performs numerical simulation processing on the thermal parameters, and obtains an output result, which includes an initial temperature field image. A noise vector is determined based on the output result, and the noise vector and the thermal parameters are combined to obtain a combined result. The combined result is then input into a trained neural network model for prediction processing to obtain a target temperature field image. The target temperature field image includes multiple target temperature values ​​of the mold. The neural network model is obtained by adversarial training based on an initial generative adversarial network and thermal parameter samples of the mold during the die-casting process. Compared with the prior art, in this solution, after obtaining the thermal parameters of the mold during the die-casting process, the thermal parameters are numerically simulated to obtain the initial temperature field image, and the image is used as a data set for model input, providing good data guidance information for subsequent model processing, and combining the noise vector with the thermal parameters to more comprehensively determine the combination result, so that the determined combination result integrates multiple aspects of information; and because the neural network model is obtained in advance based on the initial generative adversarial network and thermal parameter sample adversarial training, its training process is similar to the application process, that is, the numerical simulation processing results are used as the data set for model training, which improves the quality of the model's input data set, and then after the neural network model is trained, the combination result can be directly predicted by the neural network model, and by combining physical drive and data drive, multiple target temperature values ​​of the mold can be accurately determined, thereby improving the prediction accuracy of the mold temperature, further realizing the control of the mold temperature, and maintaining the consistency and stability of the product.

[0040] In another exemplary embodiment of the present application, in order to accurately determine multiple target temperature values ​​of the mold, adversarial training can be performed through an initial generative adversarial network to construct a neural network model, wherein the initial generative adversarial network includes a generator to be trained and a discriminator to be trained connected to the generator to be trained. Figure 3 As shown, the process of constructing the neural network model includes the following steps S301 to S305: Step S301, obtaining thermal parameter samples and noise vector samples, and performing numerical simulation processing on the thermal parameter samples to obtain simulation results.

[0041] Step S302, dividing the simulation results into a training set and a test set.

[0042] Step S303, using the training set and the noise vector samples to train the parameters in the discriminator to be trained and the generator to be trained, to obtain a generative adversarial network to be verified.

[0043] Step S304, using the test set to verify the to-be-verified generative adversarial network, and determining the to-be-verified generative adversarial network that has been successfully verified as the generative adversarial network.

[0044] Step S305, using the generator in the generative adversarial network as a neural network model.

[0045] It should be noted that the above-mentioned noise vector samples can be random noise vectors, the above-mentioned simulation results are the results obtained after numerical simulation of the thermal parameter samples, the above-mentioned training set refers to the samples used for training the initial generative adversarial network, and the test set refers to the samples used for verification of the generative adversarial network to be verified, which is used to evaluate the performance of the GAN model. The data of the test set has the same distribution characteristics as the data of the training set to ensure the validity of the evaluation.

[0046] See also Figure 4 As shown, taking the initial neural network model as the initial generative adversarial network and the proxy model as the to-be-verified generative adversarial network as an example, after obtaining the thermal parameter samples, the solver can be used to perform numerical simulation processing on the input thermal parameter samples, for example, by processing through the energy equation, so as to obtain the output parameters, which are the simulation results, and the simulation results can include the initial temperature field image samples, and then the simulation results are divided into a training set and a test set according to a preset ratio, for example, 80% of the simulation results are used as the training set, and 20% of the simulation results are used as the test set. Then the training set and the noise vector sample are input into the neural network model (initial generative adversarial network), and the parameters in the to-be-trained generator and the to-be-trained discriminator in the initial generative adversarial network are adversarially trained to obtain the to-be-verified generative adversarial network, which is the proxy model, and then the test set is input into the proxy model for verification processing, so as to obtain the generative adversarial network, and the generator in the generative adversarial network is used as the neural network model. Among them, the above-mentioned to-be-verified generative adversarial network includes a to-be-verified generator and a to-be-verified discriminator.

[0047] Specifically, in the process of training the parameters in the discriminator to be trained and the generator to be trained using the training set and the noise vector samples, part of the data set can be selected from the training set and the noise vector samples can be obtained from the noise distribution, and the noise vector samples and the thermal parameter samples can be input in parallel into the generator to be trained to obtain a simulated temperature field image; feature extraction processing is performed on part of the data set to obtain a real temperature field image; the real temperature field image and the simulated temperature field image are input into the discriminator to be trained to obtain a prediction result; a loss function is constructed according to the prediction result, and the discriminator to be trained and the generator to be trained are trained in sequence according to the minimization of the loss function to obtain a generative adversarial network to be verified. The prediction result is used to characterize the probability that the real temperature field image and the simulated temperature field image are evaluated as real samples.

[0048] See also Figure 5As shown, after randomly sampling a part of the data set x from the training set, the thermal parameter sample is used as the auxiliary variable y, and the noise vector sample is obtained from the noise distribution as the random variable z, and then the auxiliary variable y and the random variable z are combined to obtain synthetic temperature field information, and the synthetic temperature field information is input into the generator to be trained to obtain a simulated temperature field image, which is a false temperature field image; and the randomly sampled part of the data set x is feature extracted to obtain a real simulated temperature field image, and then the simulated temperature field image and the real temperature field image are input into the discriminator to be trained to obtain a prediction result, which is used to characterize the probability of whether the input image data is true.

[0049] The adversarial training process of the generator to be trained and the discriminator to be trained can be expressed by the following formula: ; Where G is the generator to be trained, D is the discriminator to be trained, x is the partial data set input to the discriminator to be trained, y is the thermal parameter sample, which is an additional information and can be any auxiliary information, and z is the noise vector sample, which can be a random variable of Gaussian distribution, as the input of the generator to be trained. x~pdata(x) [log D(x|y)] is the probability of judging that the input data is a real temperature field image, G(z|y) represents the simulated temperature field image obtained by inputting the noise vector sample and the thermal parameter sample into the generator to be trained, D(G(z|y)) represents the result obtained by inputting the simulated temperature field image generated by the generator to be trained into the discriminator to be trained, and E z~pdata(z) [log(1-D(G(z|y)))] represents the probability that the discriminator to be trained evaluates the simulated temperature field image generated by the generator as a real sample. The task of the above generator is to generate samples close to real samples, that is, to generate simulated temperature field images with similar distribution to the real temperature field images.

[0050] The above GAN model uses thermal parameter samples to generate simulated temperature field images. The above generator to be trained is denoted as: G:R Z ×R P →R T The above discriminator to be trained is denoted as D:R T ×R P →{0,1}, where the prediction result of the discriminator to be trained is between 0 and 1, Z is the dimension of the noise vector sample, P is the dimension of the thermal parameter sample, and T is the dimension of the simulated temperature field image.

[0051] In the process of inputting the noise vector samples and thermal parameter samples into the generator to be trained in parallel to obtain the simulated temperature field image, the thermal parameter sample p can be first encoded to obtain the thermal parameter encoding result, and the thermal parameter encoding result can be inserted into the noise vector sample z to obtain the generated vector, and then the generated vector is input into the generator to be trained to obtain the simulated temperature field image.

[0052] It should be noted that when encoding the thermal parameter sample, a preset encoding algorithm can be used for processing. The thermal parameter sample p is used as an additional condition to enable the generated simulated temperature field image to meet the corresponding thermal parameter sample, which can be expressed by the following formula: T = G(z,p) z∈R Z ~N(0,1), p∈R P ; Among them, z is the dimension of the noise vector sample, which can be a vector randomly drawn from the standard normal distribution N(0, 1), P is the dimension of the thermal parameter sample, and T is the dimension of the simulated temperature field image.

[0053] In this embodiment, see Figure 6 As shown, the above-mentioned generator to be trained may include multiple neural network layers, for example, multiple convolutional layers. After obtaining the thermal parameter sample and the noise vector sample z, the thermal parameter sample may be encoded to obtain a thermal parameter encoding result, and then the thermal parameter encoding result is inserted into the noise vector sample z to obtain a generated vector, and the generated vector is input into the convolutional layer of the generator for feature extraction processing, thereby generating a simulated temperature field image that meets the thermal parameter sample.

[0054] The above-mentioned discriminator to be trained may include convolutional layers, pooling layers, fully connected layers, and activation function layers. Among them, the convolutional layer is used to extract image features, such as edges, textures, etc.; the pooling layer is used to reduce the data dimension; the fully connected layer is used for the final classification judgment; the activation function layer is used to introduce nonlinear factors to enhance the expression ability of the discriminator. Common activation functions include ReLU, Leaky ReLU, etc., so that a probability value between 0 and 1 is output, indicating the probability that the input data is real data. Close to 1 means that the discriminator believes that the input data is real, and close to 0 means that it is considered to be false data generated by the generator. The prediction result output by the above-mentioned discriminator to be trained may include a first result and a second result. The real temperature field image and the simulated temperature field image are input into the discriminator to be trained. In the process of obtaining the prediction result, the real temperature field image and the simulated temperature field image are respectively input into the convolution layer in the discriminator to be trained for feature extraction to obtain the first feature information corresponding to the real temperature field image and the second feature information corresponding to the simulated temperature field image; the thermal parameter encoding result is fused with the first feature information to obtain the first fusion information, and the first fusion information is input into the fully connected layer in the discriminator to be trained for processing to obtain the first result; the thermal parameter encoding result is fused with the second feature information in parallel to obtain the second fusion information, and the second fusion information is input into the fully connected layer in the discriminator to be trained for processing to obtain the second result. Among them, the first result is the first probability value of the real temperature field image being evaluated as a real sample, and the second result is the second probability value of the simulated temperature field image being evaluated as a real sample.

[0055] See also Figure 7 As shown, the real temperature field image is input into multiple convolutional layers in the discriminator to be trained, and feature extraction is performed in each convolutional layer in turn until the first feature information is obtained after the last convolutional layer is processed, and then the thermal parameter encoding result and the first feature information are fused, which can be a vector splicing or combination method, to obtain the first fusion information, and then the first fusion information is input into the fully connected layer and the activation function layer in the discriminator to be trained for processing to obtain the first probability value of the real temperature field image being evaluated as a real sample. Similarly, the simulated temperature field image output by the generator to be trained is input into the convolutional layer, the fully connected layer and the activation function layer in the discriminator to be trained for processing in turn to obtain the second probability value of the simulated temperature field image being evaluated as a real sample.

[0056] As a feasible method, after obtaining the prediction result, a loss function is constructed according to the prediction result, and then according to the loss function minimization, an iterative algorithm is used to update the parameters in the generator to be trained and the discriminator to be trained to obtain a verified generative adversarial network; the test set is used to verify the generative adversarial network to obtain a generative adversarial network, and the generator in the generative adversarial network is used as a neural network model.

[0057] Wherein, when updating the parameters in the discriminator to be trained and the generator to be trained, it can be to update the matrix parameters such as the weight matrix and the bias matrix in the discriminator to be trained and the generator to be trained. Wherein, the above-mentioned weight matrix and bias matrix include but are not limited to the matrix parameters in each layer in the discriminator to be trained and the generator to be trained. In this embodiment, the purpose of training the initial generative adversarial network is achieved by updating the parameters in the discriminator to be trained and the generator to be trained.

[0058] The above loss function may include: a pixel loss component, a perceptual loss component, and an adversarial loss component. The pixel loss component is used to characterize the loss between the real temperature field image and the simulated temperature field image, and the perceptual loss component is used to characterize the difference between the first hidden layer feature of the real temperature field image and the second hidden layer feature of the simulated temperature field image. The first hidden layer feature is the output feature of the hidden layer of the real temperature field image in the discriminator to be trained, and the second hidden layer feature is the output feature of the hidden layer of the simulated temperature field image in the discriminator to be trained.

[0059] The above loss function loss can be expressed by the following formula: ; Among them, λ is the weight value corresponding to the pixel loss component, pixelated_loss is the pixel loss component, which is the loss between the simulated temperature field image generated at the pixel level and the real temperature field image. The purpose of the pixel loss component is to make the generated simulated temperature field image as similar as possible to the real temperature field image, including similarity and temperature range; percep_loss is the perceptual loss component, which is not only based on pixel-level comparison, but also considers the high-level features of the image; adver_loss is the adversarial loss component, which is a binary cross entropy loss function. The above pixel loss component pixelated_loss can be expressed by the following formula: ; Among them, y i is the real temperature field image, To predict temperature-temperature field image, that is, simulate temperature field image, n is the number of pixels.

[0060] The above perceptual loss component percep_loss can be expressed by the following formula: ; Among them, λ j To balance the hyperparameters affected by F different hidden layers, F represents the number of hidden layers, is the loss of the jth hidden layer. The perceptual loss component can be understood as the weighted sum of the losses of multiple hidden layers.

[0061] The above-mentioned adversarial loss component adver_loss can be expressed by the following formula: ; Among them, y i is the real temperature field image, To predict the temperature-temperature field image, that is, to simulate the temperature field image.

[0062] It is understandable that when the parameters in the initial generative adversarial network are updated by the loss function, when it is determined according to the loss function that the initial generative adversarial network has not converged, the parameters in the model are adjusted so that the initial generative adversarial network converges, thereby obtaining a generative adversarial network. The convergence of the initial generative adversarial network may mean that the difference between the simulated temperature field image generated by the generator and the real temperature field image is less than a preset threshold, or the rate of change of the difference between the simulated temperature field image generated by the generator and the real temperature field image approaches a certain lower value. When the calculated loss function is small, or the difference between the loss function output from the previous round of iteration approaches 0, it is considered that the initial generative adversarial network converges, thereby obtaining a generative adversarial network, and the generator in the generative adversarial network is used as a neural network model.

[0063] As an implementable method, in the process of updating the parameters in the discriminator to be trained and the generator to be trained, the parameters of the generator can be fixed first, and the parameters in the discriminator to be trained are updated to obtain the trained discriminator, and then the parameters of the trained discriminator are fixed, and new data is generated through the generator. The loss of the generator is calculated according to the output of the discriminator and the parameters of the generator to be trained are updated to obtain the generative adversarial network to be verified. The generative adversarial network to be verified is verified using the test set to obtain the generative adversarial network; the generator in the generative adversarial network is used as a neural network model.

[0064] Optionally, after obtaining the first result, a first loss function may be constructed based on the first result, and the first loss function may be minimized to train the discriminator to be trained to obtain the discriminator to be verified; and noise vector samples may be randomly extracted from the noise distribution, and the noise vector samples and thermal parameter samples may be input into the generator to be trained to obtain the next simulated temperature field image; the next simulated temperature field image may be input into the trained discriminator to construct a second loss function, and the second loss function may be minimized to train the generator to be trained to obtain the generator to be verified. Among them, the first loss function is the loss function when training the discriminator, which is used to measure its ability to distinguish between real temperature field images and simulated temperature field images; the second loss function is the loss function when training the generator, which is used to measure the degree to which the generated simulated temperature field image is misjudged as real data by the discriminator. When the second loss function gradually stabilizes, it means that the quality of the data generated by the generator is gradually stabilizing.

[0065] After the neural network model is trained, the thermal parameters of the mold during the die-casting process are obtained, and after numerical simulation processing is performed on it to obtain the output result, the noise vector is determined, and the thermal parameters and noise vector are input into the trained neural network model for prediction processing to obtain a target temperature field image including multiple target temperature values ​​of the mold.

[0066] In this embodiment, by obtaining thermal parameter samples and noise vector samples, and after numerically simulating the thermal parameter samples to obtain simulation results, the samples are divided into training sets and test sets, thereby improving the data set quality of the training process, and then the training set is used to train the parameters in the discriminator to be trained and the generator to be trained to obtain the generative adversarial network to be verified; and the test set is used to verify the generative adversarial network to obtain the generative adversarial network, so that the generator in the generative adversarial network can be accurately used as a neural network model, which is convenient for realizing accurate control of the mold temperature, thereby ensuring product quality, avoiding material cooling too fast or too slow, thereby reducing defects such as shrinkage, bubbles, and deformation, and ensuring product consistency and stability; at the same time, reasonable mold temperature can reduce thermal stress, prevent excessive wear or cracking of the mold, and extend the service life of the mold. Furthermore, the prediction and control of the mold temperature can shorten the production cycle, reduce the temperature adjustment time in the production process, improve the overall production efficiency, and through the accurate prediction of the mold temperature, the process flow and equipment configuration can be optimized in the design stage to improve the controllability of production.

[0067] Based on the same inventive concept, the embodiment of the present application also provides a device for realizing the mold temperature prediction of the above-mentioned one-piece thin-walled die-casting process. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme recorded in the above-mentioned method, so the specific limitations in the embodiments of one or more one-piece thin-walled die-casting process mold temperature prediction devices provided below can refer to the limitations of the one-piece thin-walled die-casting process mold temperature prediction method above, and will not be repeated here.

[0068] In an exemplary embodiment, Figure 8 As shown, a one-piece thin-walled die-casting process mold temperature prediction device is provided, comprising: An acquisition module 510 is used to acquire thermal parameters of the mold during the die-casting process; Numerical simulation module 520, used to perform numerical simulation processing on thermal parameters to obtain output results; the output results include an initial temperature field image; A combination module 530 is used to determine a noise vector according to the output result, and to combine the noise vector and the thermal parameter to obtain a combination result; The prediction module 540 is used to input the combination result into the trained neural network model for prediction processing to obtain a target temperature field image; the target temperature field image includes multiple target temperature values ​​of the mold; the neural network model is obtained by adversarial training based on the initial generative adversarial network and the thermal parameter samples of the mold during the die-casting process.

[0069] As an optional implementation, the neural network model is constructed in the following manner: Acquire thermal parameter samples and noise vector samples, and perform numerical simulation processing on the thermal parameter samples to obtain simulation results; Divide the simulation results into training set and test set; The training set and the noise vector samples are used to train the parameters of the discriminator to be trained and the generator to be trained, and a generative adversarial network to be verified is obtained; The test set is used to verify the to-be-verified generative adversarial network, and the successfully verified to-be-verified generative adversarial network is determined as the obtained generative adversarial network; The generator in the generative adversarial network is used as a neural network model.

[0070] As an optional implementation, the above device is specifically used for: Select part of the data set from the training set and obtain noise vector samples from the noise distribution; The noise vector samples and the thermal parameter samples are input into the generator to be trained in parallel to obtain a simulated temperature field image; Perform feature extraction on some data sets to obtain the real temperature field image; Inputting the real temperature field image and the simulated temperature field image into the discriminator to be trained to obtain a prediction result; the prediction result is used to characterize the probability that the real temperature field image or the simulated temperature field image is evaluated as a real sample; A loss function is constructed based on the prediction results, and according to the minimization of the loss function, the discriminator to be trained and the generator to be trained are trained in turn to obtain the generative adversarial network to be verified.

[0071] As an optional implementation, the above device is also used for: Encoding the thermal parameter samples to obtain thermal parameter encoding results; Insert the hot parameter encoding result into the noise vector sample to obtain the generated vector; The generated vector is input into the generator to be trained to obtain a simulated temperature field image.

[0072] As an optional implementation, the above device is also used for: The real temperature field image and the simulated temperature field image are respectively input into the convolution layer of the discriminator to be trained for feature extraction, so as to obtain first feature information corresponding to the real temperature field image and second feature information corresponding to the simulated temperature field image; The thermal parameter encoding result is fused with the first feature information to obtain first fused information, and the first fused information is input into the fully connected layer in the discriminator to be trained for processing to obtain a first result, which is a first probability value of the real temperature field image being evaluated as a real sample; The thermal parameter encoding result is fused with the second feature information in parallel to obtain second fused information, and the second fused information is input into the fully connected layer in the discriminator to be trained for processing to obtain a second result, which is a second probability value of the simulated temperature field image being evaluated as a real sample.

[0073] As an optional implementation, the loss function includes: a pixel loss component, a perceptual loss component, and an adversarial loss component; The pixel loss component is used to characterize the loss between the real temperature field image and the simulated temperature field image; The perceptual loss component is used to characterize the difference between the first hidden layer features of the real temperature field image and the second hidden layer features of the simulated temperature field image; the first hidden layer features are the output features of the hidden layer of the real temperature field image in the discriminator to be trained, and the second hidden layer features are the output features of the hidden layer of the simulated temperature field image in the discriminator to be trained.

[0074] As an optional embodiment, the thermal parameters include at least one of the following: thermal conductivity, specific heat capacity, latent heat of fusion, density, solidification range, solid fraction, thermal expansion coefficient, radiation coefficient, heat conduction boundary conditions and cooling rate.

[0075] Among them, the one-piece thin-walled part die-casting process mold temperature prediction device provided in the embodiment of the present application, after obtaining the thermal parameters of the mold during the die-casting process, performs numerical simulation processing on the thermal parameters to obtain the initial temperature field image, and uses it as the data set for model input, to provide good data guidance information for subsequent model processing, and combines the noise vector with the thermal parameters to more comprehensively determine the combination result, so that the determined combination result integrates multiple aspects of information; and because the neural network model is obtained in advance according to the initial generative adversarial network and thermal parameter sample adversarial training, its training process is similar to the application process, that is, the numerical simulation processing results are used as the data set for model training, which improves the quality of the model's input data set, and then after the neural network model is trained, the combination result can be directly predicted and processed by the neural network model, and by combining physical drive and data drive, multiple target temperature values ​​of the mold are accurately determined, thereby improving the prediction accuracy of the mold temperature, further realizing the control of the mold temperature, and maintaining the consistency and stability of the product.

[0076] In an exemplary embodiment, a computer device is provided. The computer device may be a server or a terminal. The internal structure diagram thereof may be as follows: Fig. 9 As shown. The computer device includes a processor, a memory, an input / output interface (Input / Output, referred to as I / O) and a communication interface. Among them, the processor, the memory and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store video tag processing data. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a mold temperature prediction method for a one-piece thin-walled die-casting process is implemented.

[0077] Those skilled in the art will understand that Fig. 9 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0078] In an exemplary embodiment, a computer device is further provided, including a memory and a processor, wherein a computer program is stored in the memory, and the processor implements the steps in the above-mentioned method embodiments when executing the computer program.

[0079] In an exemplary embodiment, a computer-readable storage medium is provided, storing a computer program, and when the computer program is executed by a processor, the steps in the above method embodiments are implemented.

[0080] In an exemplary embodiment, a computer program product is provided, including a computer program, and when the computer program is executed by a processor, the steps in the above method embodiments are implemented.

[0081] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant regulations.

[0082] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM may be in various forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM).

[0083] The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. The non-relational database may include a distributed database based on blockchain, etc., but is not limited thereto. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., but is not limited thereto.

[0084] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0085] This article uses specific examples to illustrate the principles and implementation methods of this application. The description of the above embodiments is only used to help understand the method and core ideas of this application. At the same time, for those skilled in the art, according to the ideas of this application, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting this application.

Claims

1. A method for predicting mold temperature of a die-casting process for an integrated thin-walled part, characterized in that: The mold temperature prediction method of the one-piece thin-walled part die-casting process includes: Obtain the thermal parameters of the mold during the die casting process; Performing numerical simulation processing on the thermal parameters to obtain output results; the output results include an initial temperature field image; Determine a noise vector according to the output result, and combine the noise vector and the thermal parameter to obtain a combination result; The combined result is input into a trained neural network model for prediction processing to obtain a target temperature field image; the target temperature field image includes multiple target temperature values ​​of the mold; the neural network model is obtained by adversarial training based on an initial generative adversarial network and thermal parameter samples of the mold during the die-casting process.

2. The mold temperature prediction method for die-casting process of one-piece thin-walled parts according to claim 1 is characterized in that: The initial generative adversarial network includes a generator to be trained and a discriminator to be trained connected to the generator to be trained; the neural network model is constructed in the following manner: Acquiring thermal parameter samples and noise vector samples, and performing numerical simulation processing on the thermal parameter samples to obtain simulation results; Dividing the simulation results into a training set and a test set; Using the training set and the noise vector sample to train the parameters in the discriminator to be trained and the generator to be trained, to obtain a generative adversarial network to be verified; Using the test set to verify the to-be-verified generative adversarial network, and determining the to-be-verified generative adversarial network that has been successfully verified as a generative adversarial network; The generator in the generative adversarial network is used as the neural network model.

3. The mold temperature prediction method for die-casting process of one-piece thin-walled parts according to claim 2 is characterized in that: Using the training set and the noise vector sample to train the parameters in the discriminator to be trained and the generator to be trained to obtain a generative adversarial network to be verified, including: Selecting a portion of the data set from the training set; Inputting the noise vector sample and the thermal parameter sample into the generator to be trained in parallel to obtain a simulated temperature field image; Performing feature extraction processing on the partial data set to obtain a real temperature field image; Inputting the real temperature field image and the simulated temperature field image into a discriminator to be trained to obtain a prediction result; the prediction result is used to characterize the probability that the real temperature field image or the simulated temperature field image is evaluated as a real sample; A loss function is constructed according to the prediction result, and the discriminator to be trained and the generator to be trained are trained in sequence according to minimization of the loss function to obtain the generative adversarial network to be verified.

4. The mold temperature prediction method for die-casting process of one-piece thin-walled parts according to claim 3 is characterized in that: Inputting the noise vector sample and the thermal parameter sample into the generator to be trained in parallel to obtain a simulated temperature field image, including: Performing encoding processing on the thermal parameter samples to obtain thermal parameter encoding results; Inserting the hot parameter encoding result into the noise vector sample to obtain a generated vector; The generated vector is input into the generator to be trained to obtain the simulated temperature field image.

5. The mold temperature prediction method for die-casting process of one-piece thin-walled parts according to claim 4 is characterized in that: The prediction result includes a first result and a second result; Inputting the real temperature field image and the simulated temperature field image into the discriminator to be trained to obtain a prediction result, including: Inputting the real temperature field image and the simulated temperature field image into the convolution layer of the discriminator to be trained respectively for feature extraction, so as to obtain first feature information corresponding to the real temperature field image and second feature information corresponding to the simulated temperature field image; The thermal parameter encoding result is fused with the first feature information to obtain first fused information, and the first fused information is input into the fully connected layer in the discriminator to be trained for processing to obtain the first result, which is a first probability value of the real temperature field image being evaluated as a real sample; The thermal parameter encoding result and the second feature information are fused in parallel to obtain second fused information, and the second fused information is input into the fully connected layer in the discriminator to be trained for processing to obtain the second result, which is a second probability value of the simulated temperature field image being evaluated as a real sample.

6. The mold temperature prediction method for die-casting process of one-piece thin-walled parts according to claim 5 is characterized in that: The loss function includes: a pixel loss component, a perceptual loss component and an adversarial loss component; The pixel loss component is used to characterize the loss between the real temperature field image and the simulated temperature field image; The perceptual loss component is used to characterize the difference between the first hidden layer features of the real temperature field image and the second hidden layer features of the simulated temperature field image; the first hidden layer features are the output features of the hidden layer of the real temperature field image in the discriminator to be trained, and the second hidden layer features are the output features of the hidden layer of the simulated temperature field image in the discriminator to be trained.

7. The mold temperature prediction method for die-casting process of one-piece thin-walled parts according to claim 1 is characterized in that: The thermal parameters include at least one of the following: thermal conductivity, specific heat capacity, latent heat of fusion, density, solidification range, solid fraction, thermal expansion coefficient, radiation coefficient, heat conduction boundary conditions and cooling rate.

8. An integrated thin-walled die-casting process mold temperature prediction device, characterized in that: The one-piece thin-walled part die-casting process mold temperature prediction device comprises: An acquisition module, used to acquire thermal parameters of the mold during the die-casting process; A numerical simulation module, used for performing numerical simulation processing on the thermal parameters to obtain output results; the output results include an initial temperature field image; A combination module, used for determining a noise vector according to the output result, and combining the noise vector and the thermal parameter to obtain a combination result; A prediction module is used to input the combination result into a trained neural network model for prediction processing to obtain a target temperature field image; the target temperature field image includes multiple target temperature values ​​of the mold; the neural network model is obtained by adversarial training based on an initial generative adversarial network and thermal parameter samples of the mold during the die-casting process.

9. A computer device comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the method for predicting the mold temperature of a one-piece thin-walled part die-casting process as described in any one of claims 1-7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method for predicting the mold temperature of a one-piece thin-walled part die-casting process according to any one of claims 1 to 7 are implemented.

Images