Prediction Method and Device, Medium, and Terminal for Fatigue Life of Metal Materials

Abstract

The present application discloses a method and device, medium, and terminal for predicting the fatigue life of a metal material, relating to the fields of artificial intelligence technology and materials science. The main purpose is to solve the problem that the existing data-driven prediction model for the fatigue life of materials can only meet the prediction accuracy requirements within the data range, while the prediction performance outside the data range cannot meet the needs. It includes: obtaining the current attribute parameters of the target material and the current stress parameters applied to the target material, where the target material is any metal material; based on the material fatigue life prediction model that has completed model training, performing a fatigue life prediction operation according to the current attribute parameters and the current stress parameters to obtain the fatigue life prediction result of the target material; and converting the fatigue life prediction result into the form of a fatigue life heat map for output.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence technology and the field of materials science, and particularly relates to a method and device for predicting the fatigue life of metal materials, a medium, and a terminal. Background Art

[0002] In production, fatigue failure of materials is an important cause of structural component failure. Therefore, accurately predicting the fatigue life of materials is particularly important. Accurately predicting the fatigue life of materials can not only significantly improve the reliability and safety of products, but also effectively reduce maintenance costs and extend the service life of products. Traditional methods for predicting the fatigue life of materials mainly rely on experimental tests and theoretical models based on experience. Among them, due to the time-consuming and laborious nature of experimental tests and the need for a large number of specimens, the cost is high; theoretical models based on experience are usually only applicable to specific materials and working conditions, lacking universality and flexibility, and having defects in dealing with complex working condition changes and reflecting the coupling effect of multiple factors.

[0003] With the rapid development of artificial intelligence technology, especially machine learning and deep learning, data-driven methods for predicting the fatigue life of materials have received extensive attention. Specifically, the fatigue life of materials is predicted by learning the implicit rules in a large amount of experimental data, which can overcome the limitations of traditional methods to a certain extent and improve the accuracy of prediction results.

[0004] However, due to the fact that a pure data-driven model only relies on training data to learn the associations between data, the completed data-driven model can only meet the prediction accuracy requirements within the data range, and the prediction performance outside the data range cannot meet the needs. Summary of the Invention

[0005] In view of this, the present application provides a method and device for predicting the fatigue life of metal materials, a medium, and a terminal, mainly aiming at the problem that the existing data-driven prediction model for the fatigue life of materials can only meet the prediction accuracy requirements within the data range, and the prediction performance outside the data range cannot meet the needs.

[0006] According to one aspect of the present application, a method for predicting the fatigue life of metal materials is provided, including:

[0007] Obtaining the current attribute parameters of the target material and the current stress parameters applied to the target material, where the target material is any metal material;

[0008] Based on the material fatigue life prediction model with completed model training, perform a fatigue life prediction operation according to the current attribute parameters and the current stress parameters to obtain the fatigue life prediction result of the target material. The material fatigue life prediction model is constructed based on a fully connected neural network and trained using a model loss function composed of a data-driven loss sub-function and a physical monotonicity constraint loss sub-function. The true fatigue life utilized by the data-driven loss sub-function is calculated according to the material fatigue life mathematical model. The physical monotonicity constraint loss sub-function is constructed based on the derivatives of the attribute parameters and the stress parameters with respect to the material fatigue life mathematical model. The material fatigue life mathematical model is constructed based on the crack propagation theory and the energy release theory;

[0009] Convert the fatigue life prediction result into the form of a fatigue life heat map for output.

[0010] Preferably, before the material fatigue life prediction model based on the completed model training performs a fatigue life prediction operation according to the current attribute parameters and the current stress parameters to obtain the fatigue life prediction result of the target material, the method further includes:

[0011] Construct an initial material fatigue life prediction model based on a fully connected neural network;

[0012] Construct a model loss function, which is composed of a data-driven loss sub-function and a physical monotonicity constraint loss sub-function. The model loss function is expressed by the following formula,

[0013] ,

[0014] where, represents the model loss function, represents the data-driven loss sub-function, represents a hyperparameter for controlling the weight of the physical monotonicity constraint loss sub-function, represents the physical monotonicity constraint loss sub-function;

[0015] Obtain the historical attribute parameters of the historical material and the historical stress parameters applied to the historical material. Based on the material fatigue life mathematical model, calculate the true historical fatigue life of the historical material according to the historical attribute parameters and the historical stress parameters, and generate multiple data groups based on the historical attribute parameters, the historical stress parameters, and the true historical fatigue life;

[0016] Randomly select the first preset number of data groups from the multiple groups of the data groups to generate a training data set. Based on the model loss function and the training data set, perform model training on the initial material fatigue life prediction model, record the training loss curve, and when it is monitored that the training loss curve converges, determine that the model training is completed to obtain an intermediate material fatigue life prediction model;

[0017] Randomly select the second preset number of data groups from the multiple groups of the data groups to generate a test data set. Based on the intermediate material fatigue life prediction model, perform a fatigue life prediction operation according to the historical attribute parameters and historical stress parameters in the test data set to obtain the predicted historical fatigue life of the historical material, and generate the first test result of the intermediate material fatigue life prediction model according to the error between the predicted historical fatigue life and the true historical fatigue life in the test data set;

[0018] Take each data group included in the test data set as a target data group, calculate the gradient of the predicted historical fatigue life corresponding to the target data group with respect to the historical attribute parameters and historical stress parameters in the target data group, and compare whether the direction of the gradient is consistent with the expected gradient direction. If not, mark the target data group;

[0019] Count the number of data groups with marks, calculate the proportion of the number in the number of data groups included in the test data set, and generate the second test result of the intermediate material fatigue life prediction model according to the comparison result between the proportion and the preset proportion;

[0020] If both the first test result and the second test result are qualified, determine that the intermediate material fatigue life prediction model is a material fatigue life prediction model that has completed model training.

[0021] Preferably, before constructing the model loss function, the method further includes:

[0022] Construct a data-driven loss sub-function, which is expressed by the following formula

[0023] ,

[0024] where represents the number of data groups included in the training data set, represents the true historical fatigue life calculated according to the material fatigue life mathematical model, represents the predicted historical fatigue life predicted based on the material fatigue life prediction model.

[0025] Preferably, before constructing the model loss function, the method further includes:

[0026] Construct a physical monotonicity constraint loss sub-function, and the physical monotonicity constraint loss sub-function is expressed by the following formula,

[0027] ,

[0028] wherein, represents the th input feature in the th data group in the training data set, represents a symbol factor for indicating the direction of the physical relationship, represents the derivative of the mathematical model of material fatigue life with respect to the attribute parameters and stress parameters.

[0029] Preferably, the mathematical model of material fatigue life is constructed based on the crack propagation theory and the energy release theory, and is expressed by the following formula,

[0030] ,

[0031] wherein, represents the fatigue life, C and m represent material constants, Y represents the material shape factor, represents the stress amplitude, Δ K IC represents the material fracture toughness parameter, Ra represents the surface roughness.

[0032] Preferably, before generating multiple data groups based on the historical attribute parameters, the historical stress parameters, and the true historical fatigue life, the method further includes:

[0033] Normalize the historical attribute parameters and the historical stress parameters to obtain the normalized historical attribute parameters and the normalized historical stress parameters;

[0034] Perform a logarithmic conversion on the true historical fatigue life to obtain the logarithmically converted true historical fatigue life;

[0035] Generate multiple data groups based on the normalized historical attribute parameters, the normalized historical stress parameters, and the logarithmically converted true historical fatigue life.

[0036] Preferably, the current attribute parameter is the current surface roughness parameter, and the current stress parameters include the current stress amplitude parameter and the current strain amplitude parameter.

[0037] According to another aspect of the present application, there is provided a prediction device for the fatigue life of a metal material, including:

[0038] A parameter acquisition module for acquiring the current property parameters of a target material and the current stress parameters applied to the target material, where the target material is any metal material;

[0039] A fatigue life prediction module for performing a fatigue life prediction operation based on a material fatigue life prediction model that has completed model training, according to the current property parameters and the current stress parameters, to obtain a fatigue life prediction result of the target material. The material fatigue life prediction model is constructed based on a fully connected neural network and trained using a model loss function composed of a data-driven loss sub-function and a physical monotonicity constraint loss sub-function. The true fatigue life utilized by the data-driven loss sub-function is calculated according to a material fatigue life mathematical model, and the physical monotonicity constraint loss sub-function is constructed based on the derivatives of the property parameters and stress parameters with respect to the material fatigue life mathematical model. The material fatigue life mathematical model is constructed based on the crack propagation theory and the energy release theory;

[0040] A prediction result output module for converting the fatigue life prediction result into the form of a fatigue life heat map for output.

[0041] Preferably, before the fatigue life prediction module, the device further includes a model training module for:

[0042] Constructing an initial material fatigue life prediction model based on a fully connected neural network;

[0043] Constructing a model loss function, where the model loss function is composed of a data-driven loss sub-function and a physical monotonicity constraint loss sub-function, and the model loss function is expressed by the following formula,

[0044] ,

[0045] where, represents the model loss function, represents the data-driven loss sub-function, represents a hyperparameter for controlling the weight of the physical monotonicity constraint loss sub-function, represents the physical monotonicity constraint loss sub-function;

[0046] Acquiring the historical property parameters of historical materials and the historical stress parameters applied to the historical materials, calculating the true historical fatigue life of the historical materials according to the historical property parameters and the historical stress parameters based on the material fatigue life mathematical model, and generating multiple data sets based on the historical property parameters, the historical stress parameters, and the true historical fatigue life;

[0047] Randomly select the first preset number of data groups from the multiple groups of the data groups to generate a training data set. Based on the model loss function and the training data set, train the initial material fatigue life prediction model, record the training loss curve, and determine that the model training is completed when it is monitored that the training loss curve converges, so as to obtain an intermediate material fatigue life prediction model;

[0048] Randomly select the second preset number of data groups from the multiple groups of the data groups to generate a test data set. Based on the intermediate material fatigue life prediction model, perform a fatigue life prediction operation according to the historical attribute parameters and historical stress parameters in the test data set to obtain the predicted historical fatigue life of the historical material, and generate the first test result of the intermediate material fatigue life prediction model according to the error between the predicted historical fatigue life and the true historical fatigue life in the test data set;

[0049] Take each data group included in the test data set as a target data group, calculate the gradient of the predicted historical fatigue life corresponding to the target data group with respect to the historical attribute parameters and historical stress parameters in the target data group, and compare whether the direction of the gradient is consistent with the expected gradient direction. If not, mark the target data group;

[0050] Count the number of data groups with marks, calculate the proportion of the number in the number of data groups included in the test data set, and generate the second test result of the intermediate material fatigue life prediction model according to the comparison result between the proportion and the preset proportion;

[0051] If both the first test result and the second test result are qualified, determine that the intermediate material fatigue life prediction model is the material fatigue life prediction model that has completed model training.

[0052] Preferably, the model training module is further configured to:

[0053] Construct a data-driven loss sub-function, which is expressed by the following formula,

[0054] ,

[0055] where, represents the number of data groups included in the training data set, represents the true historical fatigue life calculated according to the material fatigue life mathematical model, represents the predicted historical fatigue life predicted based on the material fatigue life prediction model.

[0056] Preferably, the model training module is further configured to:

[0057] Construct a physical monotonicity constraint loss sub-function, and the physical monotonicity constraint loss sub-function is expressed as the following formula:

[0058] ,

[0059] where, represents the th input feature in the th data group in the training data set, represents a symbol factor for indicating the direction of the physical relationship, represents the derivative of the material fatigue life mathematical model with respect to the attribute parameters and stress parameters.

[0060] Preferably, the material fatigue life mathematical model is constructed based on the crack propagation theory and the energy release theory, and is expressed as the following formula:

[0061] ,

[0062] where, represents the fatigue life, C and m represent material constants, Y represents the material shape factor, represents the stress amplitude, Δ K IC represents the material fracture toughness parameter, Ra represents the surface roughness.

[0063] Preferably, the model training module is further configured to:

[0064] Perform normalization processing on the historical attribute parameters and the historical stress parameters to obtain the normalized historical attribute parameters and the normalized historical stress parameters;

[0065] Perform logarithmic transformation processing on the true historical fatigue life to obtain the logarithmically transformed true historical fatigue life;

[0066] Generate multiple data groups based on the normalized historical attribute parameters, the normalized historical stress parameters, and the logarithmically transformed true historical fatigue life.

[0067] Preferably, the current attribute parameter is the current surface roughness parameter, and the current stress parameters include the current stress amplitude parameter and the current strain amplitude parameter.

[0068] According to another aspect of the present application, a storage medium is provided, and at least one executable instruction is stored in the storage medium, and the executable instruction causes the processor to perform the operations corresponding to the prediction method of the fatigue life of the metal material as described above.

[0069] According to another aspect of the present application, a terminal is provided, including: a processor, a memory, a communication interface, and a communication bus, and the processor, the memory, and the communication interface complete communication with each other through the communication bus;

[0070] The memory is used to store at least one executable instruction, and the executable instruction causes the processor to execute the operations corresponding to the above-mentioned prediction method for the fatigue life of metal materials.

[0071] By means of the above technical solution, the technical solution provided by the embodiments of the present application has at least the following advantages:

[0072] The present application provides a method and device, medium, and terminal for predicting the fatigue life of metal materials. First, the current attribute parameters of the target material and the current stress parameters applied to the target material are obtained, and the target material is any metal material; secondly, based on the material fatigue life prediction model that has completed model training, the fatigue life prediction operation is performed according to the current attribute parameters and the current stress parameters to obtain the fatigue life prediction result of the target material. The material fatigue life prediction model is constructed based on a fully connected neural network and is trained using a model loss function composed of a data-driven loss sub-function and a physical monotonicity constraint loss sub-function. The true fatigue life used by the data-driven loss sub-function is calculated according to the material fatigue life mathematical model, and the physical monotonicity constraint loss sub-function is constructed based on the derivatives of the attribute parameters and stress parameters of the material fatigue life mathematical model. The material fatigue life mathematical model is constructed based on the crack propagation theory and the energy release theory; finally, the fatigue life prediction result is converted into the form of a fatigue life heat map for output. Compared with the prior art, in the training process of the material fatigue life prediction model, the embodiments of the present application use a model loss function composed of a data-driven loss sub-function and a physical monotonicity constraint loss sub-function for model training, so that the model can not only accurately fit the training data, but also follow the physical laws of material fatigue behavior, thereby effectively avoiding the situation of increasing prediction result errors of traditional data-driven models outside the data range; moreover, since the physical monotonicity constraint loss sub-function is constructed based on the derivatives (i.e., gradients) of the attribute parameters and stress parameters of the material fatigue life mathematical model, it ensures that the fatigue life prediction result conforms to the expected physical relationship, enhancing the interpretability of the material fatigue life prediction model.

[0073] The above description is only an overview of the technical solution of the present application. In order to be able to understand the technical means of the present application more clearly, it can be implemented according to the content of the specification. And in order to make the above and other purposes, features, and advantages of the present application more obvious and understandable, the specific embodiments of the present application are specifically exemplified below. Description of the Drawings

[0074] By reading the following detailed description of the preferred embodiments, various other advantages and benefits will become clear to those of ordinary skill in the art. The drawings are only for the purpose of showing the preferred embodiments and are not considered to be a limitation of the present application. Moreover, throughout the drawings, the same reference numerals are used to represent the same components. In the drawings:

[0075] Figure 1 A flowchart showing a method for predicting the fatigue life of a metal material provided by an embodiment of the present application is shown;

[0076] Figure 2 A thermogram showing the fatigue life prediction of 2024-T3 aluminum alloy provided by an embodiment of the present application when the surface roughness Ra = 0.5 μm is shown;

[0077] Figure 3 A flowchart showing the training of a material fatigue life prediction model provided by an embodiment of the present application is shown;

[0078] Figure 4 A curve graph showing the training loss provided by an embodiment of the present application is shown;

[0079] Figure 5 A block diagram showing the composition of a device for predicting the fatigue life of a metal material provided by an embodiment of the present application is shown;

[0080] Figure 6 A schematic diagram showing the structure of a terminal provided by an embodiment of the present application is shown. Detailed Embodiments

[0081] Hereinafter, exemplary embodiments of the present disclosure will be described in more detail with reference to the drawings. Although the exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments set forth herein. On the contrary, these embodiments are provided so that the present disclosure can be more thoroughly understood and the scope of the present disclosure can be completely conveyed to those skilled in the art.

[0082] At the same time, it should be understood that, for the sake of description, the dimensions of the various parts shown in the drawings are not drawn in actual proportional relationship.

[0083] The following description of at least one exemplary embodiment is merely illustrative and in no way limiting of the present application and its application or use.

[0084] Technologies, methods, and devices known to those of ordinary skill in the relevant art may not be discussed in detail, but where appropriate, the technologies, methods, and devices should be regarded as part of the specification.

[0085] It should be noted that like reference numerals and letters refer to like items in the following figures, and thus, once an item is defined in one figure, it will not be further discussed in subsequent figures.

[0086] Embodiments of the present application can be applied to a computer system / server, which can operate together with many other general-purpose or special-purpose computing system environments or configurations. Examples of well-known computing systems, environments, and / or configurations suitable for use with a computer system / server include, but are not limited to: personal computer systems, server computer systems, thin clients, thick clients, handheld or laptop devices, microprocessor-based systems, set-top boxes, programmable consumer electronics, network personal computers, minicomputer systems, mainframe computer systems, and distributed cloud computing technology environments including any of the above systems, and so on.

[0087] The computer system / server can be described in the general context of computer system-executable instructions, such as program modules, executed by a computer system. Generally, program modules can include routines, programs, target programs, components, logic, data structures, etc., which perform specific tasks or implement specific abstract data types. The computer system / server can be implemented in a distributed cloud computing environment where tasks are performed by remote processing devices linked through a communication network. In a distributed cloud computing environment, program modules can be located on local or remote computing system storage media including storage devices.

[0088] Embodiments of the present application provide a method for predicting the fatigue life of a metal material, as Figure 1 shown, the method includes:

[0089] 101. Obtain the current property parameters of the target material and the current stress parameters applied to the target material.

[0090] Among them, the target material is any metal material; the current property parameters are the current surface roughness parameters, and the current stress parameters include the current stress amplitude parameter and the current strain amplitude parameter. In embodiments of the present application, the current execution end can be the material fatigue prediction module of the engineering decision-making system, and use the pre-trained material fatigue life prediction model to predict the fatigue life of the target material according to the current surface roughness parameter, the current stress amplitude parameter, and the current strain amplitude parameter of the target material.

[0091] It should be noted that since the material fatigue life prediction model is obtained by training the model using a model loss function composed of a data-driven loss function and a physical monotonicity constraint loss function, the true fatigue life utilized by the data-driven loss function is calculated based on the material fatigue life mathematical model, the physical monotonicity constraint loss function is constructed based on the derivatives of the attribute parameters and stress parameters of the material fatigue life mathematical model, and the material fatigue life mathematical model is constructed based on the crack propagation theory and the energy release theory. Since the vast majority of metallic materials follow the crack propagation theory and the energy release theory, the material fatigue life prediction model is not limited to the metallic materials used during training and can be applied to the vast majority of metallic materials, effectively improving the generalization ability and applicability of the material fatigue life mathematical model. Therefore, the target material can be any metallic material.

[0092] 102. Based on the material fatigue life prediction model with completed model training, perform a fatigue life prediction operation according to the current attribute parameters and current stress parameters to obtain the fatigue life prediction result of the target material.

[0093] Among them, the material fatigue life prediction model is constructed based on a fully connected neural network and obtained by training the model using a model loss function composed of a data-driven loss function and a physical monotonicity constraint loss function; the true fatigue life utilized by the data-driven loss function is calculated based on the material fatigue life mathematical model, the physical monotonicity constraint loss function is constructed based on the derivatives of the attribute parameters and stress parameters of the material fatigue life mathematical model, and the material fatigue life mathematical model is constructed based on the crack propagation theory and the energy release theory. In the embodiment of the present application, during the training process of the material fatigue life prediction model, the model is trained using a model loss function composed of a data-driven loss function and a physical monotonicity constraint loss function, enabling the model to not only accurately fit the training data but also follow the physical laws of material fatigue behavior, thereby effectively avoiding the situation where the prediction result error of the traditional data-driven model increases outside the data range; moreover, since the physical monotonicity constraint loss function is constructed based on the derivatives (i.e., gradients) of the attribute parameters and stress parameters of the material fatigue life mathematical model, it ensures that the fatigue life prediction result conforms to the expected physical relationship, enhancing the interpretability of the material fatigue life prediction model.

[0094] 103. Convert the fatigue life prediction result into the form of a fatigue life heat map for output.

[0095] In the embodiment of the present application, the fatigue life prediction result can be displayed through visualization means such as a fatigue life heat map to improve the intuitiveness of the display of the fatigue life prediction result. Exemplarily, Figure 2It is a heat map for predicting the fatigue life of 2024 - T3 aluminum alloy when the surface roughness Ra = 0.5μm.

[0096] Compared with the prior art, in the training process of the material fatigue life prediction model in the embodiments of the present application, a model loss function composed of a data - driven loss sub - function and a physical monotonicity constraint loss sub - function is used for model training. This enables the model to not only accurately fit the training data but also follow the physical laws of material fatigue behavior, thus effectively avoiding the situation where the prediction error of the traditional data - driven model increases outside the data range. Moreover, since the physical monotonicity constraint loss sub - function is constructed based on the derivatives (i.e., gradients) of the material fatigue life mathematical model with respect to the property parameters and stress parameters, it ensures that the fatigue life prediction results conform to the expected physical relationships, enhancing the interpretability of the material fatigue life prediction model.

[0097] In one embodiment of the present application, for further limitation and illustration, as Figure 3 shown, before step 102 of the embodiment performs the fatigue life prediction operation based on the material fatigue life prediction model that has completed model training and obtains the fatigue life prediction result of the target material according to the current property parameters and the current stress parameters, the method of the embodiment further includes:

[0098] 201. Construct an initial material fatigue life prediction model based on a fully - connected neural network.

[0099] In the embodiments of the present application, the structure of the initial material fatigue life prediction model can consist of an input layer, three hidden layers, and an output layer. Among them, the number of neurons in the hidden layers can be configured according to the training data set. For example, hidden layer 1 contains 64 neurons, hidden layer 2 contains 64 neurons, and hidden layer 3 contains 32 neurons, and the ReLU function is used as the activation function to enhance the non - linear expression ability of the material fatigue life prediction model. Further, the output layer can use a linear activation function to adapt to the prediction of continuous values of fatigue life. The structure of this initial material fatigue life prediction model can effectively

[0100] capture the complex non - linear relationship between the property parameters of the material, the current stress parameters applied to the material, and the fatigue life, thereby improving the accuracy of the model prediction results.

[0101] 202. Construct a data - driven loss sub - function.

[0102] Specifically, the data - driven loss sub - function is expressed as the following formula,

[0103] ,

[0104] where Indicates the number of data groups included in the training data set, Indicates the true historical fatigue life calculated according to the material fatigue life mathematical model, Indicates the predicted historical fatigue life predicted based on the material fatigue life prediction model.

[0105] It should be noted that the error between the true historical fatigue life and the predicted historical fatigue life is calculated by the data-driven loss sub-function, which can ensure the fitting accuracy of the model on the training data.

[0106] 203. Construct a physical monotonicity constraint loss sub-function.

[0107] Specifically, the physical monotonicity constraint loss sub-function is expressed as the following formula,

[0108] ,

[0109] Among them, Indicates the th input feature in the th data group in the training data set, Indicates the sign factor (-1 or 1) used to indicate the direction of the physical relationship, Indicates the derivative of the material fatigue life mathematical model with respect to the attribute parameters and stress parameters. When Indicates the stress amplitude parameter When it increases, the fatigue life Indicates the stress amplitude parameter should decrease. When Indicates the strain amplitude parameter When it increases, the fatigue life should decrease. When Indicates the strain amplitude parameter When it increases, the fatigue life should decrease. When Indicates the surface roughness parameter Ra, Indicates that when the surface roughness parameter Ra increases, the fatigue life should decrease, Indicates the material fatigue life mathematical model.

[0110] It should be noted that since the physical monotonicity constraint loss sub-function is constructed based on the derivative (i.e., gradient) of the material fatigue life mathematical model with respect to the attribute parameters and stress parameters, it ensures that the fatigue life prediction result conforms to the expected physical relationship, enhancing the interpretability of the material fatigue life prediction model.

[0111] 204. Construct a model loss function.

[0112] Specifically, the model loss function consists of a data-driven loss sub-function and a physical monotonicity constraint loss sub-function. The model loss function is expressed as the following formula:

[0113] ,

[0114] where, represents the model loss function, represents the data-driven loss sub-function, represents a hyperparameter for controlling the weight of the physical monotonicity constraint loss sub-function, represents the physical monotonicity constraint loss sub-function.

[0115] It should be noted that the larger the value of the hyperparameter , the stronger the physical monotonicity constraint, but the computational cost will increase accordingly. Therefore, the value of the hyperparameter needs to be configured in combination with the actual computing power.

[0116] Optionally, during the model training process, the Adam optimizer can be used for model training. Exemplarily, the learning rate can be set to 0.001 to minimize the total loss function. During the training process, the batch size can be set to 32, and the number of training epochs can be set to 100. In each training epoch, the loss value is calculated and accumulated, and the training loss curve is recorded to monitor the convergence of the model, thereby improving the model training efficiency.

[0117] It should be noted that in the embodiments of the present application, a model loss function is constructed using a data-driven loss sub-function and a physical monotonicity constraint loss sub-function, so that the material fatigue life prediction model trained based on this model loss function can not only accurately fit the training data, but also follow the physical laws of material fatigue behavior, thereby effectively avoiding the situation where the prediction result error of the traditional data-driven model increases outside the data range.

[0118] 205. Obtain the historical attribute parameters of historical materials and the historical stress parameters applied to the historical materials. Based on the material fatigue life mathematical model, calculate the true historical fatigue life of the historical materials according to the historical attribute parameters and the historical stress parameters, and generate multiple data groups based on the historical attribute parameters, the historical stress parameters, and the true historical fatigue life.

[0119] Among them, the historical materials can be any metal material. Specifically, the material fatigue life mathematical model is constructed based on the crack propagation theory and the energy release theory, and is expressed as the following formula:

[0120] ,

[0121] where, represents the fatigue life, Cand m represents a material constant Y represents a material shape factor represents the stress amplitude, Δ K IC represents a material fracture toughness parameter Ra represents the surface roughness

[0122] In the embodiments of the present application, the data set includes historical attribute parameters, historical stress parameters, and corresponding true historical fatigue lives

[0123] Optionally, before generating multiple data sets based on the historical attribute parameters, historical stress parameters, and true historical fatigue lives in step 205 of the embodiment, the method of the embodiment further includes: normalizing the historical attribute parameters and historical stress parameters to obtain the normalized historical attribute parameters and the normalized historical stress parameters; performing a logarithmic conversion process on the true historical fatigue life to obtain the logarithmically converted true historical fatigue life; generating multiple data sets based on the normalized historical attribute parameters, the normalized historical stress parameters, and the logarithmically converted true historical fatigue life

[0124] It should be noted that through the normalization process, the historical attribute parameters and historical stress parameters can be converted into values between [0, 1], so that all parameters are at the same order of magnitude, while reducing the consumed computing resources; through the logarithmic conversion process, the value of the true historical fatigue life can be reduced, so that the computing resources are reduced, and at the same time, the correlation between variables is not affected

[0125] 206. Randomly select a first preset number of data sets from multiple data sets to generate a training data set. Based on the model loss function and the training data set, perform model training on the initial material fatigue life prediction model, record the training loss curve, and when it is monitored that the training loss curve converges, determine that the model training is completed to obtain an intermediate material fatigue life prediction model

[0126] Among them, the first preset number can be 70% of the number of data sets, and a training data set is generated for performing model training on the initial material fatigue life prediction model. In the embodiments of the present application, the model training process can be displayed through visualization means such as a training loss curve graph, etc., to improve the intuitiveness of the display of the model training process, such as Figure 4 as shown, by recording the training loss curve to display the convergence of the model loss function during the model training process, and when it is monitored that the training loss curve converges, determine that the model training is completed to obtain an intermediate material fatigue life prediction model

[0127] 207. Randomly select a second preset number of data groups from multiple groups of data groups to generate a test data set. Based on the intermediate material fatigue life prediction model, perform fatigue life prediction operations according to the historical attribute parameters and historical stress parameters in the test data set to obtain the predicted historical fatigue life of the historical material, and generate the first test result of the intermediate material fatigue life prediction model according to the error between the predicted historical fatigue life and the true historical fatigue life in the test data set.

[0128] Among them, the second preset number can be 30% of the number of data groups, and a test data set is generated for testing the intermediate material fatigue life prediction model. First, based on the intermediate material fatigue life prediction model, perform fatigue life prediction operations according to the historical attribute parameters and historical stress parameters in the test data set to obtain the predicted historical fatigue life of the historical material, and then calculate the mean square error (MSE) or mean absolute percentage error (MAPE) between the predicted historical fatigue life and the true historical fatigue life in the test data set, and generate the first test result of the intermediate material fatigue life prediction model. Specifically, if the error is less than the preset error threshold, the first test result is qualified; if the error is greater than the preset error threshold, the first test result is unqualified.

[0129] 208. Take each data group included in the test data set as a target data group, calculate the gradient of the predicted historical fatigue life corresponding to the target data group with respect to the historical attribute parameters and historical stress parameters in the target data group, and compare whether the direction of the gradient is consistent with the expected gradient direction. If not, mark the target data group.

[0130] In the embodiments of the present application, first calculate the gradient of the predicted historical fatigue life corresponding to the target data group with respect to the historical surface roughness parameter, historical stress amplitude parameter, and historical strain amplitude parameter in the target data group. Further, compare whether the direction of the calculated gradient is consistent with the expected gradient direction, where the expected gradient direction is that when the surface roughness parameter increases, the fatigue life should decrease; when the stress amplitude parameter increases, the fatigue life should decrease; when the strain amplitude parameter increases, the fatigue life should decrease, and at the same time mark the inconsistent data groups.

[0131] 209. Count the number of data groups with marks, calculate the proportion of the number in the number of data groups included in the test data set, and generate the second test result of the intermediate material fatigue life prediction model according to the comparison result between the proportion and the preset proportion.

[0132] In the embodiments of the present application, the proportion of the number of data groups with inconsistent statistical directions in the number of data groups included in the test data set is calculated. If this proportion exceeds the preset proportion, the second test result is unqualified; if this proportion does not exceed the preset proportion, the second test result is qualified. Here, the preset proportion can be set to 10% based on experience.

[0133] 210. If both the first test result and the second test result are qualified, it is determined that the intermediate material fatigue life prediction model is a material fatigue life prediction model that has completed model training.

[0134] In the embodiments of the present application, only when both the first test result and the second test result are qualified, it is determined that the prediction accuracy of the intermediate material fatigue life prediction model meets the requirements. At this time, it is determined that the intermediate material fatigue life prediction model is a material fatigue life prediction model that has completed model training.

[0135] The present application provides a method for predicting the fatigue life of a metal material. First, the current attribute parameters of the target material and the current stress parameters applied to the target material are obtained, where the target material is any metal material. Secondly, based on the material fatigue life prediction model that has completed model training, fatigue life prediction operations are performed according to the current attribute parameters and the current stress parameters to obtain the fatigue life prediction result of the target material. The material fatigue life prediction model is constructed based on a fully connected neural network and trained using a model loss function composed of a data-driven loss sub-function and a physical monotonicity constraint loss sub-function. The true fatigue life used by the data-driven loss sub-function is calculated according to the material fatigue life mathematical model, and the physical monotonicity constraint loss sub-function is constructed based on the derivatives of the attribute parameters and stress parameters with respect to the material fatigue life mathematical model. The material fatigue life mathematical model is constructed based on the crack propagation theory and the energy release theory. Finally, the fatigue life prediction result is converted into the form of a fatigue life heat map for output. Compared with the prior art, in the embodiments of the present application, during the training process of the material fatigue life prediction model, a model loss function composed of a data-driven loss sub-function and a physical monotonicity constraint loss sub-function is used for model training, enabling the model to not only accurately fit the training data but also follow the physical laws of material fatigue behavior, thus effectively avoiding the increase in prediction result errors of traditional data-driven models outside the data range. Moreover, since the physical monotonicity constraint loss sub-function is constructed based on the derivatives (i.e., gradients) of the attribute parameters and stress parameters with respect to the material fatigue life mathematical model, it ensures that the fatigue life prediction result conforms to the expected physical relationship, enhancing the interpretability of the material fatigue life prediction model.

[0136] Further, as for the above Figure 1For the implementation of the method described above, an embodiment of the present application provides a device for predicting the fatigue life of a metal material, as Figure 5 shown, the device includes:

[0137] A parameter acquisition module 31, a fatigue life prediction module 32, and a prediction result output module 33;

[0138] The parameter acquisition module 31 is configured to acquire the current property parameters of the target material and the current stress parameters applied to the target material, where the target material is any metal material;

[0139] The fatigue life prediction module 32 is configured to perform a fatigue life prediction operation based on a material fatigue life prediction model that has completed model training, according to the current property parameters and the current stress parameters, to obtain a fatigue life prediction result of the target material. The material fatigue life prediction model is constructed based on a fully connected neural network and is trained using a model loss function composed of a data-driven loss sub-function and a physical monotonicity constraint loss sub-function. The true fatigue life used by the data-driven loss sub-function is calculated according to a material fatigue life mathematical model, and the physical monotonicity constraint loss sub-function is constructed based on the derivatives of the property parameters and stress parameters with respect to the material fatigue life mathematical model. The material fatigue life mathematical model is constructed based on the crack growth theory and the energy release theory;

[0140] The prediction result output module 33 is configured to output the fatigue life prediction result in the form of a fatigue life heat map.

[0141] In a specific application scenario, before the fatigue life prediction module, the device further includes a model training module, which is configured to:

[0142] Construct an initial material fatigue life prediction model based on a fully connected neural network;

[0143] Construct a model loss function, which is composed of a data-driven loss sub-function and a physical monotonicity constraint loss sub-function. The model loss function is expressed as the following formula,

[0144] ,

[0145] where, represents the model loss function, represents the data-driven loss sub-function, represents a hyperparameter for controlling the weight of the physical monotonicity constraint loss sub-function, represents the physical monotonicity constraint loss sub-function;

[0146] Obtain the historical attribute parameters of the historical material and the historical stress parameters applied to the historical material. Based on the mathematical model of the material fatigue life, calculate the true historical fatigue life of the historical material according to the historical attribute parameters and the historical stress parameters, and generate multiple data sets based on the historical attribute parameters, the historical stress parameters, and the true historical fatigue life;

[0147] Randomly select a first preset number of data sets from multiple groups of the data sets to generate a training data set. Based on the model loss function and the training data set, perform model training on the initial material fatigue life prediction model, record the training loss curve, and determine that the model training is completed when it is monitored that the training loss curve converges, and obtain an intermediate material fatigue life prediction model;

[0148] Randomly select a second preset number of data sets from multiple groups of the data sets to generate a test data set. Based on the intermediate material fatigue life prediction model, perform a fatigue life prediction operation according to the historical attribute parameters and historical stress parameters in the test data set to obtain the predicted historical fatigue life of the historical material, and generate a first test result of the intermediate material fatigue life prediction model according to the error between the predicted historical fatigue life and the true historical fatigue life in the test data set;

[0149] Take each data set included in the test data set as a target data set, calculate the gradient of the predicted historical fatigue life corresponding to the target data set with respect to the historical attribute parameters and historical stress parameters in the target data set, and compare whether the direction of the gradient is consistent with the expected gradient direction. If not, mark the target data set;

[0150] Count the number of data sets with marks, calculate the proportion of the number in the number of data sets included in the test data set, and generate a second test result of the intermediate material fatigue life prediction model according to the comparison result between the proportion and the preset proportion;

[0151] If both the first test result and the second test result are qualified, determine the intermediate material fatigue life prediction model as the material fatigue life prediction model that has completed model training.

[0152] In a specific application scenario, the model training module is further configured to:

[0153] Construct a data-driven loss sub-function, and the data-driven loss sub-function is expressed by the following formula,

[0154] ,

[0155] where, Indicates the number of data groups included in the training data set, Indicates the true historical fatigue life calculated according to the material fatigue life mathematical model, Indicates the predicted historical fatigue life predicted based on the material fatigue life prediction model.

[0156] In a specific application scenario, the model training module is further configured to:

[0157] Construct a physical monotonicity constraint loss sub-function, and the physical monotonicity constraint loss sub-function is expressed as the following formula,

[0158] ,

[0159] where, Indicates the th input feature in the th data group in the training data set, Indicates the sign factor for indicating the direction of the physical relationship, Indicates the derivative of the material fatigue life mathematical model with respect to the attribute parameters and the stress parameters.

[0160] In a specific application scenario, the material fatigue life mathematical model is constructed based on the crack propagation theory and the energy release theory, and is expressed as the following formula,

[0161] ,

[0162] where, Indicates the fatigue life, C and m Indicates material constants, Y Indicates the material shape factor, Indicates the stress amplitude, Δ K IC Indicates the material fracture toughness parameter, Ra Indicates the surface roughness.

[0163] In a specific application scenario, the model training module is further configured to:

[0164] Normalize the historical attribute parameters and the historical stress parameters to obtain the normalized historical attribute parameters and the normalized historical stress parameters;

[0165] Perform logarithmic transformation on the true historical fatigue life to obtain the logarithmically transformed true historical fatigue life;

[0166] Generate a plurality of data groups based on the normalized historical attribute parameters, the normalized historical stress parameters, and the logarithmically transformed true historical fatigue life.

[0167] In a specific application scenario, the current attribute parameter is the current surface roughness parameter, and the current stress parameter includes the current stress amplitude parameter and the current strain amplitude parameter.

[0168] The present application provides a device for predicting the fatigue life of a metal material. First, the current attribute parameter of the target material and the current stress parameter applied to the target material are obtained, where the target material is any metal material. Secondly, based on a material fatigue life prediction model that has completed model training, a fatigue life prediction operation is performed according to the current attribute parameter and the current stress parameter to obtain the fatigue life prediction result of the target material. The material fatigue life prediction model is constructed based on a fully connected neural network and is trained using a model loss function composed of a data-driven loss sub-function and a physical monotonicity constraint loss sub-function. The true fatigue life used by the data-driven loss sub-function is calculated according to a material fatigue life mathematical model, and the physical monotonicity constraint loss sub-function is constructed based on the derivatives of the attribute parameter and the stress parameter with respect to the material fatigue life mathematical model. The material fatigue life mathematical model is constructed based on the crack propagation theory and the energy release theory. Finally, the fatigue life prediction result is converted into the form of a fatigue life heat map for output. Compared with the prior art, in the training process of the material fatigue life prediction model in the embodiment of the present application, a model loss function composed of a data-driven loss sub-function and a physical monotonicity constraint loss sub-function is used for model training, so that the model can not only accurately fit the training data, but also follow the physical laws of the material fatigue behavior, thereby effectively avoiding the situation where the prediction result error of the traditional data-driven model increases outside the data range. And, since the physical monotonicity constraint loss sub-function is constructed based on the derivatives (i.e., gradients) of the attribute parameter and the stress parameter with respect to the material fatigue life mathematical model, it ensures that the fatigue life prediction result conforms to the expected physical relationship, enhancing the interpretability of the material fatigue life prediction model.

[0169] According to an embodiment of the present application, a storage medium is provided. The storage medium stores at least one executable instruction, and the computer executable instruction can execute the method for predicting the fatigue life of a metal material in any of the above method embodiments.

[0170] Based on such an understanding, the technical solution of the present application can be embodied in the form of a software product. The software product can be stored in a non-volatile storage medium (which can be a CD-ROM, a USB flash drive, a mobile hard disk, etc.), and includes several instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in various implementation scenarios of the present application.

[0171] Figure 6 The figure shows a schematic structural diagram of a terminal provided according to an embodiment of the present application. The specific embodiments of the present application do not limit the specific implementation of the terminal.

[0172] As Figure 6 shown, the terminal may include: a processor 402, a communications interface 404, a memory 406, and a communication bus 408.

[0173] Among them: the processor 402, the communications interface 404, and the memory 406 communicate with each other through the communication bus 408.

[0174] The communications interface 404 is used to communicate with network elements of other devices such as clients or other servers.

[0175] The processor 402 is used to execute the program 410, and specifically can execute the relevant steps in the embodiment of the method for predicting the fatigue life of the above-mentioned metal material.

[0176] Specifically, the program 410 may include program code, and the program code includes computer operation instructions.

[0177] The processor 402 may be a central processing unit CPU, or a specific integrated circuit ASIC (Application Specific Integrated Circuit), or one or more integrated circuits configured to implement the embodiments of the present application. One or more processors included in the computer device may be of the same type of processor, such as one or more CPUs; or may be of different types of processors, such as one or more CPUs and one or more ASICs.

[0178] The memory 406 is used to store the program 410. The memory 406 may include high-speed RAM memory, and may also include non-volatile memory, such as at least one disk memory.

[0179] The program 410 is specifically used to cause the processor 402 to perform the following operations:

[0180] Obtain the current property parameters of the target material and the current stress parameters applied to the target material, where the target material is any metal material;

[0181] Based on the material fatigue life prediction model with completed model training, perform a fatigue life prediction operation according to the current attribute parameters and the current stress parameters to obtain the fatigue life prediction result of the target material. The material fatigue life prediction model is constructed based on a fully connected neural network and trained using a model loss function composed of a data-driven loss sub-function and a physical monotonicity constraint loss sub-function. The true fatigue life utilized by the data-driven loss sub-function is calculated according to the material fatigue life mathematical model. The physical monotonicity constraint loss sub-function is constructed based on the derivatives of the attribute parameters and stress parameters with respect to the material fatigue life mathematical model. The material fatigue life mathematical model is constructed based on the crack propagation theory and the energy release theory;

[0182] Convert the fatigue life prediction result into the form of a fatigue life heat map for output.

[0183] The storage medium may also include an operating system and a network communication module. The operating system is a program that manages the hardware and software resources of the physical device for the above-mentioned method for predicting the fatigue life of metal materials, and supports the operation of the information processing program and other software and / or programs. The network communication module is used to implement communication between components within the storage medium and communication between other hardware and software in the information processing physical device.

[0184] Each embodiment in this specification is described in a progressive manner. The key point of each embodiment is to illustrate the differences from other embodiments. For the same or similar parts between each embodiment, reference can be made to each other. For the system embodiment, since it basically corresponds to the method embodiment, the description is relatively simple, and reference can be made to the corresponding part of the method embodiment for the relevant content.

[0185] The methods and systems of the present application can be implemented in many ways. For example, the methods and systems of the present application can be implemented through software, hardware, firmware, or any combination of software, hardware, and firmware. The above order of the steps for the method is only for illustration, and the steps of the method of the present application are not limited to the specific order described above, unless otherwise specifically stated. In addition, in some embodiments, the present application can also be implemented as a program recorded in a recording medium, and these programs include machine-readable instructions for implementing the method according to the present application. Therefore, the present application also covers a recording medium storing a program for executing the method according to the present application.

[0186] Obviously, those skilled in the art should understand that the various modules or steps of the present application described above can be implemented by a general-purpose computing device. They can be concentrated on a single computing device or distributed on a network composed of multiple computing devices. Optionally, they can be implemented by program codes executable by the computing device. Thus, they can be stored in a storage device and executed by the computing device. And in some cases, the steps shown or described can be executed in a sequence different from that here, or they can be separately fabricated into individual integrated circuit modules, or multiple modules or steps among them can be fabricated into a single integrated circuit module for implementation. In this way, the present application is not limited to any specific combination of hardware and software.

[0187] The foregoing is only a preferred embodiment of the present application and is not intended to limit the present application. For those skilled in the art, the present application can have various changes and modifications. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included within the protection scope of the present application.

Claims

1. A method for predicting the fatigue life of a metal material, characterized in that, Including: Obtaining the current property parameters of the target material and the current stress parameters applied to the target material, where the target material is any metal material; Based on a material fatigue life prediction model with completed model training, performing a fatigue life prediction operation according to the current property parameters and the current stress parameters to obtain the fatigue life prediction result of the target material. The material fatigue life prediction model is constructed based on a fully connected neural network and trained using a model loss function composed of a data-driven loss sub-function and a physical monotonicity constraint loss sub-function. The true fatigue life utilized by the data-driven loss sub-function is calculated according to a material fatigue life mathematical model, and the physical monotonicity constraint loss sub-function is constructed based on the derivatives of the property parameters and stress parameters of the material fatigue life mathematical model. The material fatigue life mathematical model is constructed based on the crack propagation theory and the energy release theory; Converting the fatigue life prediction result into the form of a fatigue life heat map for output; The physical monotonicity constraint loss sub-function is expressed as the following formula, where x k,i represents the k-th input feature in the i-th data group of the training dataset, and s k represents the symbol factor used to indicate the direction of the physical relationship, represents the derivative of the mathematical model of material fatigue life with respect to the attribute parameters and stress parameters; The material fatigue life mathematical model is expressed as the following formula, Among them, N f represents the fatigue life, C and m represent material constants, Y represents the material shape factor, σ represents the stress amplitude, ΔK IC represents the material fracture toughness parameter, and Ra represents the surface roughness.

2. The method according to claim 1, wherein Before the step of, based on the material fatigue life prediction model with completed model training, performing a fatigue life prediction operation according to the current property parameters and the current stress parameters to obtain the fatigue life prediction result of the target material, the method further includes: Constructing an initial material fatigue life prediction model based on a fully connected neural network; Constructing a model loss function, where the model loss function is composed of a data-driven loss sub-function and a physical monotonicity constraint loss sub-function, and the model loss function is expressed as the following formula, Loss total = Loss data + β·Loss phy , Among them, Loss total represents the model loss function, Loss data represents the data-driven loss sub-function, β represents the hyperparameter for controlling the weight of the physical monotonicity constraint loss sub-function, and Loss phy represents the physical monotonicity constraint loss sub-function; Obtaining the historical property parameters of the historical material and the historical stress parameters applied to the historical material, calculating the true historical fatigue life of the historical material according to the historical property parameters and the historical stress parameters based on the material fatigue life mathematical model, and generating multiple data groups based on the historical property parameters, the historical stress parameters, and the true historical fatigue life; Randomly extracting a first preset number of data groups from the multiple data groups to generate a training data set, training the initial material fatigue life prediction model based on the model loss function and the training data set, recording the training loss curve, and determining that the model training is completed and obtaining an intermediate material fatigue life prediction model when it is monitored that the training loss curve converges; Randomly extracting a second preset number of data groups from the multiple data groups to generate a test data set, performing a fatigue life prediction operation according to the historical property parameters and historical stress parameters in the test data set based on the intermediate material fatigue life prediction model to obtain the predicted historical fatigue life of the historical material, and generating a first test result of the intermediate material fatigue life prediction model according to the error between the predicted historical fatigue life and the true historical fatigue life in the test data set; Taking each data group included in the test data set as the target data group, calculating the gradients of the predicted historical fatigue life corresponding to the target data group with respect to the historical attribute parameters and historical stress parameters in the target data group, and comparing whether the direction of the gradient is consistent with the expected gradient direction. If not, marking the target data group; Counting the number of data groups with marks, calculating the proportion of the number in the number of data groups included in the test data set, and generating the second test result of the intermediate material fatigue life prediction model according to the comparison result between the proportion and the preset proportion; If both the first test result and the second test result are qualified, determining that the intermediate material fatigue life prediction model is a material fatigue life prediction model that has completed model training; 3. The method according to claim 2, wherein Before constructing the model loss function, the method further includes: Constructing a data-driven loss sub-function, which is expressed by the following formula Among them, n represents the number of data groups included in the training data set. represents the true historical fatigue life calculated according to the mathematical model of material fatigue life, N f,i represents the predicted historical fatigue life predicted based on the material fatigue life prediction model.

4. The method according to claim 2, wherein Before generating multiple data groups based on the historical attribute parameters, the historical stress parameters, and the true historical fatigue life, the method further includes: Performing normalization processing on the historical attribute parameters and the historical stress parameters to obtain the normalized historical attribute parameters and the normalized historical stress parameters; Performing logarithmic transformation processing on the true historical fatigue life to obtain the logarithmically transformed true historical fatigue life; Generating multiple data groups based on the normalized historical attribute parameters, the normalized historical stress parameters, and the logarithmically transformed true historical fatigue life; 5. The method according to claim 1, wherein The current attribute parameter is the current surface roughness parameter, and the current stress parameters include the current stress amplitude parameter and the current strain amplitude parameter; 6. A prediction device for the fatigue life of a metal material, characterized in that, including: A parameter acquisition module for acquiring the current attribute parameters of the target material and the current stress parameters applied to the target material, where the target material is any metal material; A fatigue life prediction module for performing a fatigue life prediction operation based on the material fatigue life prediction model that has completed model training, and obtaining the fatigue life prediction result of the target material according to the current attribute parameters and the current stress parameters. The material fatigue life prediction model is constructed based on a fully connected neural network and trained using a model loss function composed of a data-driven loss sub-function and a physical monotonicity constraint loss sub-function. The true fatigue life used by the data-driven loss sub-function is calculated according to the material fatigue life mathematical model, and the physical monotonicity constraint loss sub-function is constructed based on the derivatives of the attribute parameters and stress parameters of the material fatigue life mathematical model. The material fatigue life mathematical model is constructed based on the crack propagation theory and the energy release theory; A prediction result output module for outputting the fatigue life prediction result in the form of a fatigue life heat map; The physical monotonicity constraint loss sub-function is expressed by the following formula where x k,i represents the k-th input feature in the i-th data group of the training dataset, and s k represents a symbol factor used to indicate the direction of the physical relationship, represents the derivative of the mathematical model of material fatigue life with respect to the attribute parameters and stress parameters; The material fatigue life mathematical model is expressed by the following formula Among them, N f represents the fatigue life, C and m represent material constants, Y represents the material shape factor, σ represents the stress amplitude, and ΔK IC represents the material fracture toughness parameter, and Ra represents the surface roughness.

7. A storage medium storing at least one executable instruction, characterized in that, The executable instructions cause the processor to perform operations corresponding to the method for predicting the fatigue life of a metal material according to any one of claims 1-5.

8. A terminal, comprising: A processor, a memory, a communication interface, and a communication bus, wherein the processor, the memory, and the communication interface complete communication with each other through the communication bus; The memory is used for storing at least one executable instruction, wherein the executable instructions cause the processor to perform operations corresponding to the method for predicting the fatigue life of a metal material according to any one of claims 1-5.

Images