Simulation data generation method for metal material multi-mode fusion

Abstract

The invention belongs to the field of material data, and discloses a simulation data generation method for metal material multi-mode fusion. Comprising the steps of collecting data and performing feature extraction to obtain feature vectors; through multi-modal method fusion, a metal material multi-modal feature training data set is obtained; and constructing a metal material simulation data diffusion model and a diffusion model loss function conforming to physical condition constraints, training the diffusion model, and outputting high-quality simulation metal material data conforming to the basic physical law of the material. According to the simulation data generation method for multi-modal fusion of the metal material provided by the invention, physical constraints are introduced, so that the generated simulation data meet experimental data from a statistical angle and meet a basic physical rule; the simulation data of the metal material can be effectively generated under the condition of limited data, and the experiment and calculation cost can be remarkably reduced; and meanwhile, numerical modal data and image modal data are fused, so that the structure-performance evolution relationship of the material is better met.

Description

Technical Field

[0001] The present invention relates to the field of material data, and in particular to a method for generating simulation data of multi-modal fusion of metal materials. Background Art

[0002] Metal materials are the most important structural materials for the development of human society and play an irreplaceable role in industrial fields such as aviation, aerospace, navigation, and automobiles. Achieving high strength and high plasticity of metal materials is of great significance to the overall performance, service safety, and service life of parts.

[0003] In recent years, additive manufacturing (AM) technology has been widely used in aerospace, automobile manufacturing, biomedicine and other fields. Its notable feature is that it can manufacture parts with complex geometric structures and material properties. However, the process involves multi-scale and multi-physics coupling problems of materials, such as microstructure evolution, stress deformation, defect formation, etc. In addition, the microstructure of the material, such as grain size, porosity, grain boundary distribution, etc., directly affects its mechanical properties. These phenomena greatly affect the mechanical properties and service life of the final product. Traditional material performance testing methods are usually time-consuming and costly. Through data generation, the reliance on a large number of experiments can be reduced, thus saving time and resources, which is of great significance for the development of new materials and the improvement of existing materials. In order to improve the ability to predict the behavior of metal materials, simulation technology has gradually become a key tool.

[0004] Traditional simulation methods mainly rely on experimental and numerical simulation methods such as finite element analysis (FEA) and molecular dynamics (MD). However, these methods often require a large amount of experimental data for verification when dealing with complex, multi-defect, and multi-working metal materials, and the calculation cost is high. In addition, some parameters in the physical model (such as the evolution mechanism of microscopic defects and organizational structure) are difficult to obtain directly through experimental data, resulting in insufficient simulation accuracy.

[0005] In recent years, with the development of machine learning and deep learning technologies, data-driven methods have gradually been applied to the field of materials. By using the microstructure images, process parameters and material performance data of materials to build a data generation model, high-quality material performance data can be generated. The generated data can support simulation methods or machine learning methods for further research and analysis. However, these generative models have a series of problems in practical applications. The microstructure images, process parameters and material performance data of materials belong to different modal data, and it is a challenge to effectively integrate image data features with numerical data features. Existing generative models are mostly trained based on the statistical characteristics of data, and the generated data results lack constraints on physical laws. They may be close to experimental data in a statistical sense, but their rationality in physical behavior cannot be guaranteed.

[0006] Therefore, how to generate high-precision simulation data that conforms to physical laws and can integrate multimodal features under limited data conditions has become an urgent problem to be solved. Summary of the invention

[0007] In view of the above defects or improvement needs of the prior art, the present invention provides a method for generating simulation data of metal materials by multimodal fusion. The method constructs a multimodal data generation framework that conforms to physical laws by fusing metal material microstructure image data, material manufacturing process parameters and material performance data. The framework combines the diffusion model and the deep learning model driven by physical constraints, and introduces the physical condition constraint loss function in the training process to generate high-quality and physically consistent simulation data.

[0008] To achieve the above object, according to one aspect of the present invention, a method for generating simulation data of multi-modal fusion of metal materials is provided, comprising the following steps:

[0009] S1: Collect the process parameters, mechanical properties data and microstructure image data of the original metal materials;

[0010] S2: extracting features of the metal material microstructure image collected in step S1 through a convolutional neural network to obtain a feature vector of the metal material microstructure image;

[0011] S3: preprocessing the original metal material process parameters and mechanical property data collected in step S1 to obtain numerical feature vectors of metal material process parameters and mechanical properties;

[0012] S4: fusing the metal material microstructure image feature vector obtained in step S2 with the process parameter and performance numerical feature vector obtained in step S3 through a multimodal method to obtain a metal material multimodal feature training data set;

[0013] S5: constructing a metal material simulation data diffusion model based on the metal material multimodal feature training data set obtained in step S4;

[0014] S6: Based on the initial metal material simulation data diffusion model obtained in step S5, a diffusion model loss function that meets physical condition constraints is constructed;

[0015] S7: Based on the metal material multimodal feature training data set, the initial diffusion model and its loss function obtained in steps S4-S6, the diffusion model is trained to output high-quality simulated metal material data that conforms to the basic physical laws of the material.

[0016] Preferably, step S2 specifically includes the following steps:

[0017] S2.1: Construct a convolutional neural network model M img ,The model includes convolutional layers, pooling layers and activation functions, removing its final fully connected layer and retaining some of the feature extraction functions;

[0018] S2.2: Preprocess the image data to meet the input requirements of the convolutional neural network, and obtain the preprocessed image data set I′;

[0019] S2.3: Input the preprocessed image set I′ into the model M img In the process, the image features are extracted from low level to high level layer by layer through multiple convolution layers, and the nonlinearity is improved through activation function. Finally, the dimension and calculation amount of the feature map are reduced through pooling operation, while the important features of the image are retained to complete the extraction of the feature vector F of the microstructure image of the metal material. img .

[0020] Preferably, in step S3, the preprocessing of the process parameters and mechanical property data of the original metal material is performed by normalization, and the normalization method is:

[0021]

[0022] Among them, X represents the process parameters and mechanical properties data of the original metal material, X data Represents the numerical training data samples of metal materials.

[0023] Preferably, the fusion method in step S4 is: by introducing a gating mechanism to add dynamic weights to different modal data to adapt to multimodal feature input, retain important features and control the selective expression of features, while introducing a residual structure to improve the stability and robustness of the model, thereby completing the fusion of multimodal data features.

[0024] Preferably, step S4 specifically includes the following steps:

[0025] S4.1: Adjusting the numerical feature vector F by repetition and expansion data With the image feature vector F img to the same dimension;

[0026] S4.2: Define a gating mechanism to dynamically adjust the fusion ratio of image features and numerical features, and define a gating tensor The gating weights are calculated by a feedforward neural network as follows:

[0027] G=σ(W g ·F img +b g )

[0028] Among them, Wg is the weight matrix, bg is the bias, and σ is the activation function Sigmoid, which ensures that the value of G is between 0 and 1;

[0029] S4.3: Control the fusion weights of image and numerical modality features through the gated tensor G to form the gated fusion feature F gate :

[0030] F gate =G⊙F img +(1-G)⊙F data

[0031] S4.4: Based on the gated fusion feature, add the residual feature connection F res , and finally get the fusion feature F fusion :

[0032] F fusion =F gate (F img ,F data )+F res (F img ,F data )

[0033] Among them, the residual connection F res (F img ,F data )=F img +F data .

[0034] Preferably, the diffusion model in step S5 includes a forward diffusion process and a reverse diffusion process. The forward diffusion process maps the input metal material multimodal training data samples to the Gaussian noise space, and gradually adds noise to the data to convert it into pure Gaussian noise; the reverse diffusion module gradually removes noise from the noise, reconstructs the data and generates new metal material simulation data.

[0035] Preferably, step S5 specifically includes the following steps:

[0036] S5.1: In the forward process of the diffusion model, noise is gradually added to the real data through time step t to construct a Gaussian noise distribution, so that the training data gradually transforms from real data samples to pure noise. The diffusion step is defined as follows:

[0037]

[0038] Among them, x t is the data after adding noise, x0 is the original real metal material data, α t is a hyperparameter that controls the noise level based on the time step t, ∈~N(0,I) is the random noise sampled from the standard normal distribution;

[0039] S5.2: During the back diffusion process, the model gradually restores Gaussian noise to simulation data that conforms to the characteristics of metal materials. For each time step t, the model defines the conditional probability p θ (x t-1 |x t ), represents the denoising process from time step t to t-1; by learning the mean function μ θ (x t , t) and variance parameter σ θ (t), realize gradual denoising and generate samples closer to real data, as shown in the following formula:

[0040] x t-1 =μ θ (x,t)+σ θ (t)∈

[0041] Among them, μ θ (x t , t) predicts the data position at the current time step, σ θ (t) controls the randomness of denoising, ∈~N(0,I) is the random noise sampled from the standard normal distribution.

[0042] As a preference, the loss function Loss in step S6 total The construction is achieved by combining data loss and physical consistency as follows:

[0043] Loss total =Loss data +λ·Loss phys

[0044] Among them, data loss Loss data The evaluation index used for regression tasks is used to measure the fitting loss between the generated data and the real data; the physical consistency loss Loss phys The consistency loss between the generated data and the expected physical behavior is calculated based on the physical model and constraints; λ is the balance coefficient, which adjusts the direct weight of data loss and physical loss.

[0045] As a preference, the loss function Loss in step S6 total The construction method specifically includes the following steps:

[0046] S6.1: Calculate data loss. The diffusion model involves modeling the noise distribution of randomly sampled variables, so the expected form of the mean squared error (MSE) evaluation indicator is used to optimize the fit between the model output and the true distribution:

[0047]

[0048] in, represents the expected value of the average error under the joint distribution of sample x0 and random sampling variable t; x0 represents the real distribution data, Represents the model output data, Represents the squared Euclidean distance between the predicted value and the true value, and calculates the model loss.

[0049] S6.2: Calculate the physical consistency loss, calculate the deviation of the generated data according to the physical model, force the generated data to meet the material characteristics, and ensure that the data meets the physical laws:

[0050] Loss phys =λ JC Loss JC +λ vM Loss vM +λ HP Loss HP

[0051] Among them, λ JC , vM , HP is a weight coefficient used to balance the contribution of different physical loss items in the total loss. JC 、Loss vM 、Loss HP It is a loss function constrained by different physical conditions. The size of the loss value is proportional to the gap between the generated data and the actual physical laws.

[0052] Loss JC The loss function is defined as the resulting stress σ gen The absolute value of the difference from the Johnson-Cook model prediction is as follows:

[0053]

[0054] Among them, σ gen is the generated stress, and the right side is the stress prediction value of the Johnson-Cook model. A, B, C, m, and n are material constants. T m is the melting point of the material, and T is set to room temperature 25°C. This loss function will guide the model to generate the real material stress under different temperatures, strain rates and strain conditions;

[0055] Loss vM The loss function is defined as the equivalent stress σ calculated from the generated stress data v The yield strength of the material should not exceed yeild , as follows:

[0056] Loss yield =max(0,σ v -σ y )

[0057] in, σ1, σ2, and σ3 are the maximum stresses generated by the model on the material in different directions; the equivalent stress σ v and yield strength σ y The difference is compared with 0. When the generated equivalent stress exceeds the yield strength, the loss function is positive. If the generated data meets the yield criterion, the loss value is 0.

[0058] Loss vM The loss function is defined as the yield strength σ generated by the stress data y The absolute value of the Hall-Petch relationship varies with the grain size d, as shown below:

[0059] Loss HP =|σ y,gen -(σ0+k y d -1 / 2 )|

[0060] Among them, σ y，gen is the generated yield strength, σ0 and k y is the material constant for the strengthening effect at the grain boundary.

[0061] Preferably, the training of the diffusion model in step S7 includes initializing parameters and iterative training, and specifically includes the following steps:

[0062] S7.1: Initialize the parameters of the diffusion model and gradually adjust them during the training process according to the influence of the physical consistency loss to balance the physical consistency of the generated data and the data fitting effect;

[0063] S7.2: Iteratively train the diffusion model. For each training sample, each training data sample is converted into noise x through the forward diffusion process. t ; Generate data x' through the reverse diffusion process and calculate the loss function Loss total ; Through the back-propagation algorithm, the model parameters are updated to minimize the loss until the loss function converges and no longer decreases; by repeating the above steps, the diffusion model is trained so that it gradually learns to recover high-quality simulation data that conforms to the physical properties of metal materials from the noisy data.

[0064] In general, the above technical solution conceived by the present invention has the following beneficial effects compared with the prior art:

[0065] 1. The present invention adopts a multimodal data fusion method, which can not only conform to the numerical regression analysis results, but also conform to the material's organization-performance evolution relationship after fusing the microstructure image of the metal material, and significantly improve the quality of the generated simulation data.

[0066] 2. The present invention introduces physical constraints so that the generated simulation data not only matches the experimental data from a statistical perspective, but also satisfies basic physical laws.

[0067] 3. The method of the present invention has high generalization ability, can effectively generate simulation data of metal materials under limited data conditions, and can significantly reduce experimental and computational costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0068] Figure 1 It is a schematic diagram of the basic steps of the simulation data generation method of multi-modal fusion of metal materials of the present invention;

[0069] Figure 2 The figure is a schematic diagram of a preferred implementation process of the present invention. DETAILED DESCRIPTION

[0070] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention. In addition, the technical features involved in the various embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0071] The Diffusion Model is an emerging deep generative model. Its essence is similar to the process of training a neural network to learn image denoising to achieve a step-by-step generation process from a known Gaussian distribution to a target distribution. The core idea is to gradually add noise to the real data through a forward diffusion process, and then train a backward diffusion process to learn how to recover the data from the noise. The generation process of the diffusion model can be described as gradually denoising in multiple time steps, and finally generating new data with a distribution similar to that of the training data. The diffusion model can naturally describe the continuous dynamic process of data in its stochastic differential equation (SDE) form, and is suitable for processing the smoothness and continuity of data in regression problems. It gradually transforms the distribution of target data from a simple Gaussian distribution to the target distribution, thereby generating the corresponding regression prediction value.

[0072] Combination Figure 1 and Figure 2 This embodiment provides a method for generating simulation data of multi-modal fusion of metal materials, comprising the following steps:

[0073] Step S1: Collecting process parameters, mechanical property numerical data and microstructure image data of the original metal material.

[0074] The specific method is as follows: Use scanning electron microscope (SEM), metallographic microscope and other technologies to collect microstructure images of metal material samples. The microstructure images need to include key information such as grain structure, phase interface, defects, etc.; record the process parameters of the metal material sample process, including laser power, scanning rate, scanning spacing, layer thickness, and measure the sample material performance data through mechanical testing methods such as universal testing machine, including elastic modulus, yield strength, tensile strength, and elongation. Collect 108 metal material sample data including microstructure images, manufacturing process parameters, and mechanical properties.

[0075] Step S2: Based on the microstructure image data of the data acquisition module, the features of the metal material microstructure image are extracted through a convolutional neural network (CNN) to obtain a feature vector of the metal material microstructure image.

[0076] The specific implementation method is as follows:

[0077] Step S2.1: Construct a convolutional neural network model. Select the ResNet-50 pre-trained convolutional neural network model as the image feature extractor, set as model M img ,The model includes convolution layers, pooling layers and activation functions, removing the last fully connected layer and retaining some ,functions of feature extraction.

[0078] Step S2.2 preprocesses the image data to meet the input requirements of the convolutional neural network. First, the metal material microstructure image dataset The size is uniformly adjusted to 224×224, the number of channels is 3, and then the pixel value of each pixel point in each channel on the image is divided by 255 using the following formula, and the pixel value is normalized to the interval [0, 1] to obtain the preprocessed image dataset I′:

[0079]

[0080] Step S2.3: Input the preprocessed image set I′ into the model M img In the process of extracting microstructure image features,

[0081] Each convolution kernel slides and scans the entire image, and identifies local features by weighting and extracting features in local areas. Then, by stacking 6 convolution layers, information from low-level features (such as edges and corners) to high-level features (such as grain structure, phase interface, defects) is extracted layer by layer. After the last convolution layer, the image features are nonlinearly fused through the ReLU activation function, and then global average pooling (GAP) is performed to obtain a fixed-size image feature vector F. img , the feature extraction process is expressed as:

[0082] F img = GAP(M(I′))

[0083] The model removes the fully connected layer used for classification, and the feature vector dimension output for a single sample is C×M×N, where C is the number of channels, the depth of the feature map; M is the height of the feature map; and N is the width of the feature map. Specifically, a feature tensor of size 2048×1×1 is extracted based on the ResNet-50 convolutional neural network.

[0084] Step S3: The collected original metal material process parameters and mechanical properties are processed by a normalization method to extract numerical data feature vectors of the same scalar.

[0085] The numerical data of metal material process parameters and mechanical properties are defined as X, as shown in the following table.

[0086]

[0087] The feature extraction method of the process parameters and mechanical properties data of metal materials adopts the maximum and minimum normalization method. Specifically, the training sample X of each metal material numerical data is extracted by the following method: data :

[0088]

[0089] Among the 108 data, there are 19 types of materials. The 19 material names are converted into 22-dimensional binary numerical feature vectors through one-hot encoding, such as "17-4PH" is converted to [1, 0, 0, ..., 0], and "Ti6-Al4-V" is converted to [0, 1, 0, ..., 0];

[0090] X data Combined with the unique hot encoding feature of the material name, we get the numerical feature vector F data .

[0091] Specifically, after numerical feature extraction, the feature vector is obtained The details are shown in the following table.

[0092]

[0093] Step S4: The extracted metal material microstructure image feature vectors are integrated with the numerical feature vectors of process parameters and mechanical properties to obtain multimodal metal material training data input.

[0094] The fusion method is to add dynamic weights to different modal data by introducing a gating mechanism to adapt to multimodal feature input, retain important features and control the selective expression of features. At the same time, a residual structure is introduced to improve the stability and robustness of the model, thereby completing the fusion of multimodal data features.

[0095] That is, the dimension of the image features extracted in the previous step is much higher than the dimension of the numerical features. Direct fusion will cause the image features to dominate, greatly weaken the influence of the numerical features, and cause the loss of important information. Therefore, a gate structure and a residual structure are designed here to introduce dynamic weights to adapt to the multi-modal feature input, retain important features and control the selective expression of features, while improving the stability and robustness of the model. Specifically, the feature vector extracted from the microstructure image of the metal material is obtained from the above steps. Feature vectors extracted from process parameters and mechanical properties Furthermore, the feature vector dimension is adjusted to make the two modal data have the same shape; the influence of different modal features is dynamically adjusted through the gating mechanism, and finally the features are fused to form a multimodal feature F fusion .

[0096] Step S4.1 Adjust the numerical feature F data With image feature F img To the same dimension. Adjust the numerical features F by repeating and expanding img and F data The dimensions and shape are as follows:

[0097]

[0098] Step S4.2 defines a gating mechanism to dynamically adjust the fusion ratio of image features and numerical features. Specifically, define the gating tensor The gating weights are calculated by a feedforward neural network as follows:

[0099] G=σ(W g ·F img +b g )

[0100] Where Wg is the weight matrix, bg is the bias, and σ is the activation function Sigmoid, which ensures that the value of G is between 0 and 1.

[0101] Step S4.3 weights the fusion of image features and numerical features through the gated tensor. Specifically, the gated tensor G is used to control the fusion weights of the image and numerical modal features to form the gated fusion feature F gate , as follows:

[0102] F gate =G⊙F img +(1-G)⊙F data

[0103] Step S4.4 introduces the residual structure to form the final fusion feature. Specifically, based on the gated fusion feature, the residual feature connection F is added res, and finally get the fusion feature F fusion , as follows:

[0104] F fusion =F gate (F img ,F data )+F res (F img ,F data )

[0105] The residual connection F res (F img ,F data )=F img +F data .

[0106] Step S5: constructing a metal material simulation data diffusion model based on the metal material multimodal feature training data set.

[0107] The construction of the metal material simulation data diffusion model mainly includes the forward diffusion process and the reverse diffusion process. The forward diffusion process maps the input metal material multimodal training data samples to the Gaussian noise space, gradually adds noise to the data to transform it into pure Gaussian noise; the reverse diffusion module gradually removes noise from the noise, reconstructs the data to generate new metal material simulation data.

[0108] Step S5.1 trains the diffusion model to add noise to the real data through the forward diffusion process. Specifically, in the forward process of the diffusion model, the noise is gradually added to the real data through the time step t to construct a Gaussian noise distribution, so that the training data gradually transforms from the real data sample to pure noise. The original real metal material data obtained by the previous step is defined as x0, and the diffusion step is defined as follows:

[0109]

[0110] Among them, x t is the data after adding noise, α t is a hyperparameter that controls the noise level based on the time step t, and ∈~N(0,I) is random noise sampled from a standard normal distribution.

[0111] Step S5.2 enables the model to learn to reconstruct data from noise through the back diffusion process. In the back diffusion process, the model gradually restores Gaussian noise to simulated data that conforms to the characteristics of metal materials. This process restores the noise sample to a sample that conforms to the real distribution through the denoising step. Specifically, for each time step t, the model defines the conditional probability p θ (x t-1 |x t ), represents the denoising process from time step t to t-1. By learning the mean function μθ (x t , t) and variance parameter σ θ (t), realize gradual denoising and generate samples closer to real data, as shown in the following formula:

[0112] x t-1 =μ θ (x,t)+σ θ (t)∈

[0113] Among them, μ θ (x t , t) predicts the data position at the current time step, σ θ (t) controls the randomness of denoising, ∈~N(0,I) is the random noise sampled from the standard normal distribution.

[0114] Step S6 constructs a diffusion model loss function that meets physical condition constraints.

[0115] The present invention introduces a physical consistency loss term into the loss function of the diffusion model to ensure that the generated data conforms to the basic physical laws of metal materials. The total loss function is defined to include the data loss part and the physical consistency loss part, and the comprehensive loss function is designed as follows:

[0116] Loss total =Loss data +λ·Loss phys

[0117] Among them, Loss data For data loss, Loss phys is the physical consistency loss, and λ is the balancing coefficient for adjusting the direct weights of data loss and physical loss.

[0118] Step S6.1 calculates the data loss to measure the difference between the generated data and the real data. Specifically, the diffusion model involves modeling the noise distribution of random sampling variables, so the mean square error (MSE) evaluation index in the expected form is used to optimize the fit between the model output and the real distribution. The Euclidean distance between the feature vector of the multimodal real training data sample of the metal material and the feature vector of the predicted data is calculated, and the expected value is taken after squaring to form the data loss function:

[0119]

[0120] in, represents the expected value of the average error under the joint distribution of sample x0 and random sampling variable t; x0 represents the real distribution data, Represents the model output data, Represents the squared Euclidean distance between the predicted value and the true value.

[0121] Step S6.2 calculates the physical consistency loss to ensure the physical rationality of the generated data. The deviation of the generated data is calculated according to the physical model, and the generated data is forced to meet the material properties to ensure that the data meets the physical laws, including the Johnson-Cook model, the von Mises yield criterion, and the Hall-Petch relationship, as shown in the following formula:

[0122] Loss phys =λ JC Loss JC +λ vM Loss vM +λ HP Loss HP

[0123] Among them, λ JC , vM , HP is a weight coefficient used to balance the contribution of different physical loss items in the total loss. JC 、Loss vM 、Loss HP It is a loss function constrained by different physical conditions. The size of the loss value is proportional to the gap between the generated data and the actual physical laws.

[0124] Specifically, Loss JC The loss function is defined as the resulting stress σ gen The absolute value of the difference from the value predicted by the Johnson-Cook model. The generated data contains stress and strain. The Johnson-Cook model describes the stress response of the material at different strain rates, strains, and temperatures as follows:

[0125]

[0126] Among them, σ gen is the generated stress, and the right side is the stress prediction value of the Johnson-Cook model. A, B, C, m, and n are material constants. T m is the melting point of the material, and T is set to room temperature 25°C. This loss function will guide the model to generate realistic material stresses under different temperatures, strain rates, and strain conditions.

[0127] Loss vM The loss function is defined as the equivalent stress σ calculated from the generated stress data v The yield strength of the material should not exceed yeild The von Mises criterion is used to determine the yield condition under multiaxial stress state, as shown in the following formula:

[0128] Loss yield=max(0,σ v -σ y )

[0129] in, σ1, σ2, and σ3 are the maximum stresses generated by the model on the material in different directions. v and yield strength σ y The difference is compared with 0. When the generated equivalent stress exceeds the yield strength, the loss function is positive. If the generated data meets the yield criterion, the loss value is 0. This loss constrains the yield state of the generated data under multiaxial stress to ensure that the generated data meets the yield behavior of the material.

[0130] Loss vM The loss function is defined as the yield strength σ generated by the stress data y The absolute value of the Hall-Petch relationship should vary with the grain size d. The Hall-Petch relationship describes the effect of grain size on the yield strength of the material as follows:

[0131] Loss HP =|σ y,gen -(σ0+k y d -1 / 2 )|

[0132] Among them, σ y，gen is the generated yield strength, σ0 and k y is the material constant for the strengthening effect at the grain boundaries. This penalty term ensures that the generated data is physically consistent with the yield strength when the grain size varies.

[0133] Step S7 trains the diffusion model and outputs high-quality simulated metal material data that conforms to the basic physical laws of the material.

[0134] Model training includes initialization parameters and iterative training. Initialization parameters are used to initialize and adjust the key hyperparameters of the diffusion model at the beginning of training to ensure effective training and convergence of the model. The iterative training process includes forward diffusion to destroy data, backward diffusion to generate data, and backward propagation to optimize the model. Through the forward diffusion process, each training data sample is converted into a noise x t ; Generate data x' through the reverse diffusion process and calculate the loss function Loss total ; Update the model parameters to minimize the loss through the back propagation algorithm until the loss function converges and no longer decreases. By repeating the above steps, the diffusion model is trained to gradually learn to recover high-quality simulation data that conforms to the physical properties of metal materials from the noisy data.

[0135] Step S7.1: Initialize the parameters of the diffusion model. Initialize and adjust the key hyperparameters, set the number of diffusion steps T to 200 steps, set the noise intensity parameter α t =α0(1-decay_rate) t , where α0 is set to 0.95, decay_rate is set to 0.005, and noise is gradually added with time step t; the learning rate η is set to 0.0001; the initial value of the constraint weight of the physical loss is set to 1, and it is gradually adjusted during the training process according to the influence of the physical consistency loss to balance the physical consistency of the generated data and the data fitting effect.

[0136] Step S7.2 iteratively trains the diffusion model. Specifically, for each training sample, each training data sample is converted into noise x by the forward diffusion process. t ; Generate data x' through the reverse diffusion process and calculate the loss function Loss total ; Update the model parameters to minimize the loss through the back propagation algorithm until the loss function converges and no longer decreases. By repeating the above steps, the diffusion model is trained to gradually learn to recover high-quality simulation data that conforms to the physical properties of metal materials from the noisy data.

[0137] It will be easily understood by those skilled in the art that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for generating simulation data of multi-modal fusion of metal materials, characterized in that: The following steps are involved: S1: Collect the process parameters, mechanical properties data and microstructure image data of the original metal materials; S2: extracting features of the metal material microstructure image collected in step S1 through a convolutional neural network to obtain a feature vector of the metal material microstructure image; S3: preprocessing the original metal material process parameters and mechanical property data collected in step S1 to obtain numerical feature vectors of metal material process parameters and mechanical properties; S4: fusing the metal material microstructure image feature vector obtained in step S2 with the process parameter and performance numerical feature vector obtained in step S3 through a multimodal method to obtain a metal material multimodal feature training data set; S5: constructing a metal material simulation data diffusion model based on the metal material multimodal feature training data set obtained in step S4; S6: Based on the initial metal material simulation data diffusion model obtained in step S5, a diffusion model loss function that meets physical condition constraints is constructed; S7: Based on the metal material multimodal feature training data set, the initial diffusion model and its loss function obtained in steps S4-S6, the diffusion model is trained to output high-quality simulated metal material data that conforms to the basic physical laws of the material.

2. A method for generating simulation data for multi-modal fusion of metal materials according to claim 1, characterized in that: Step S2 specifically includes the following steps: S2.1: Construct a convolutional neural network model M img ,The model includes convolutional layers, pooling layers and activation functions, removing its final fully connected layer and retaining some of the feature extraction functions; S2.2: Preprocess the image data to meet the input requirements of the convolutional neural network, and obtain the preprocessed image data set I′; S2.3: Input the preprocessed image set I′ into the model M img In the process, the image features are extracted from low level to high level layer by layer through multiple convolution layers, and the nonlinearity is improved through activation function. Finally, the dimension and calculation amount of the feature map are reduced through pooling operation, while the important features of the image are retained to complete the extraction of the feature vector F of the microstructure image of the metal material. img .

3. The method for generating simulation data of multi-modal fusion of metal materials according to claim 1, characterized in that: In step S3, the preprocessing of the process parameters and mechanical property data of the original metal material adopts normalization processing, and the method of the normalization processing is: Among them, X represents the process parameters and mechanical properties data of the original metal material, X data Represents the numerical training data samples of metal materials.

4. The method for generating simulation data of multi-modal fusion of metal materials according to claim 1, characterized in that: The fusion method in step S4 is: by introducing a gating mechanism to add dynamic weights to different modal data to adapt to multimodal feature input, important features are retained and the selective expression of features is controlled. At the same time, a residual structure is introduced to improve the stability and robustness of the model, thereby completing the fusion of multimodal data features.

5. The method for generating simulation data of multi-modal fusion of metal materials according to claim 4, characterized in that: Step S4 specifically includes the following steps: S4.1: Adjusting the numerical feature vector F by repetition and expansion data With the image feature vector F img to the same dimension; S4.2: Define a gating mechanism to dynamically adjust the fusion ratio of image features and numerical features, and define a gating tensor The gating weights are calculated by a feedforward neural network as follows: G=σ(W g ·F img +b g ) Among them, Wg is the weight matrix, bg is the bias, and σ is the activation function Sigmoid, which ensures that the value of G is between 0 and 1; S4.3: Control the fusion weights of image and numerical modality features through the gated tensor G to form the gated fusion feature F gate : F gate =G⊙F img +(1-G)⊙F data S4.4: Based on the gated fusion feature, add the residual feature connection F res , and finally get the fusion feature F fusion : F fusion =F gate (F img ,F data )+F res (F img ,F data ) Among them, the residual connection F res (F img ,F data )=F img +F data .

6. The method for generating simulation data of multi-modal fusion of metal materials according to claim 1, characterized in that: The diffusion model in step S5 includes a forward diffusion process and a reverse diffusion process. The forward diffusion process maps the input metal material multimodal training data samples to the Gaussian noise space, and gradually adds noise to the data to convert it into pure Gaussian noise; the reverse diffusion module gradually removes noise from the noise, reconstructs the data and generates new metal material simulation data.

7. A method for generating simulation data of multi-modal fusion of metal materials as claimed in claim 6, characterized in that: Step S5 specifically includes the following steps: S5.1: In the forward process of the diffusion model, noise is gradually added to the real data through time step t to construct a Gaussian noise distribution, so that the training data gradually transforms from real data samples to pure noise. The diffusion step is defined as follows: Among them, x t is the data after adding noise, x0 is the original real metal material data, α t is a hyperparameter that controls the noise level based on the time step t, ∈~N(0,I) is the random noise sampled from the standard normal distribution; S5.2: During the back diffusion process, the model gradually restores Gaussian noise to simulation data that conforms to the characteristics of metal materials. For each time step t, the model defines the conditional probability p θ (x t-1 |x t ), represents the denoising process from time step t to t-1; by learning the mean function μ θ (x t , t) and variance parameter σ θ (t), realize gradual denoising and generate samples closer to real data, as shown in the following formula: x t-1 =μ θ (x,t)+σ θ (t)∈ Among them, μ θ (x t , t) predicts the data position at the current time step, σ θ (t) controls the randomness of denoising, ∈~N(0,I) is the random noise sampled from the standard normal distribution.

8. The method for generating simulation data of multi-modal fusion of metal materials according to claim 1, characterized in that: Loss function in step S6 total The construction is achieved by combining data loss and physical consistency as follows: Loss total =Loss data +λ·Loss phys Among them, data loss Loss data The evaluation index used for regression tasks is used to measure the fitting loss between the generated data and the real data; the physical consistency loss Loss phys The consistency loss between generated data and expected physical behavior is calculated based on the physical model and constraints; λ is the balance coefficient, which adjusts the direct weight of data loss and physical loss.

9. The method for generating simulation data of multi-modal fusion of metal materials according to claim 1, characterized in that: Loss function in step S6 total The construction method specifically includes the following steps: S6.1: Calculate data loss. The diffusion model involves modeling the noise distribution of randomly sampled variables, so the expected form of the mean squared error evaluation indicator is used to optimize the fit between the model output and the true distribution: in, represents the expected value of the average error under the joint distribution of sample x0 and random sampling variable t; x0 represents the real distribution data, Represents the model output data, Represents the squared Euclidean distance between the predicted value and the true value, and calculates the model loss; S6.2: Calculate the physical consistency loss, calculate the deviation of the generated data according to the physical model, force the generated data to meet the material characteristics, and ensure that the data meets the physical laws: Loss phys =λ JC Loss JC +λ vM Loss vM +λ HP Loss HP Among them, λ JC , vM , HP is a weight coefficient used to balance the contribution of different physical loss items in the total loss. JC 、Loss vM 、Loss HP It is a loss function constrained by different physical conditions. The size of the loss value is proportional to the gap between the generated data and the actual physical laws. Loss JC The loss function is defined as the resulting stress σ gen The absolute value of the difference from the Johnson-Cook model prediction is as follows: Among them, σ gen is the generated stress, and the right side is the stress prediction value of the Johnson-Cook model. A, B, C, m, and n are material constants. T m is the melting point of the material, and T is set to room temperature 25°C; this loss function will guide the model to generate the real material stress under different temperatures, strain rates and strain conditions; Loss vM The loss function is defined as the equivalent stress σ calculated from the generated stress data v The yield strength of the material should not exceed yeild , as follows: Loss yield =max(0,σ v -s y ) in, σ1, σ2, and σ3 are the maximum stresses generated by the model on the material in different directions; the equivalent stress σ v and yield strength σ y The difference is compared with 0. When the generated equivalent stress exceeds the yield strength, the loss function is positive. If the generated data meets the yield criterion, the loss value is 0. Loss vM The loss function is defined as the yield strength σ generated by the stress data y The absolute value of the Hall-Petch relationship varies with the grain size d, as shown below: Loss HP =|σ y,gen -(σ0+k y d -1 / 2 )| Among them, σ y，gen is the generated yield strength, σ0 and k y is the material constant for the strengthening effect at the grain boundary.

10. The method for generating simulation data of multi-modal fusion of metal materials according to claim 1, characterized in that: Training the diffusion model in step S7 includes initializing parameters and iterative training, which specifically includes the following steps: S7.1: Initialize the parameters of the diffusion model and gradually adjust them during the training process according to the influence of the physical consistency loss to balance the physical consistency of the generated data and the data fitting effect; S7.2: Iteratively train the diffusion model. For each training sample, each training data sample is converted into noise x through the forward diffusion process. t ; Generate data x' through the reverse diffusion process and calculate the loss function Loss total ; Through the back-propagation algorithm, the model parameters are updated to minimize the loss until the loss function converges and no longer decreases; by repeating the above steps, the diffusion model is trained so that it gradually learns to recover high-quality simulation data that conforms to the physical properties of metal materials from the noisy data.

Images