Plastic failure fracture prediction method and system

Abstract

The embodiment of the invention provides a plastic failure fracture prediction method and system. The method comprises the steps of training a target prediction model according to a target feature data set; according to the target prediction model, generating a target plastic failure fracture prediction result; wherein the target feature data set comprises target node features and target edge features; the target prediction model comprises a target map neural network layer, a target multi-head map attention network layer and a target long and short time network layer. In this way, calculation resource consumption in the plastic failure fracture prediction process can be reduced, the calculation process is simplified, and the plastic failure fracture simulation and iteration efficiency is improved.

Description

Technical Field

[0001] Embodiments of the present application relate to the technical field of engineering mechanics simulation, and particularly to a method and system for predicting plastic failure fracture. Background Art

[0002] Plastic Fracture refers to the fracture phenomenon that occurs after significant plastic deformation of materials. The process of plastic fracture involves complex micro-damage evolution and non-linear large deformation. The research on plastic fracture is crucial in the fields of safety assessment of engineering structures, aerospace, automobile manufacturing, etc.

[0003] In related technologies, there are problems such as large consumption of computing resources, complex calculation process, and long simulation and iteration time in plastic fracture prediction. Summary of the Invention

[0004] According to embodiments of the present application, a method and system for predicting plastic failure fracture are provided, which can reduce the consumption of computing resources in the process of plastic failure fracture prediction, simplify the calculation process, and improve the simulation and iteration efficiency of plastic failure fracture.

[0005] In the first aspect of the present application, a method for predicting plastic failure fracture is provided, including: Training a target prediction model according to a target feature dataset; Generating a target plastic failure fracture prediction result according to the target prediction model; Wherein, the target feature dataset includes: target node features and target edge features; The target prediction model includes: a target graph neural network layer, a target multi-head graph attention network layer, and a target long short-term network layer.

[0006] In some feasible embodiments, the above-mentioned training of the target prediction model according to the target feature dataset includes: Training the target graph neural network layer to perform a target feature projection operation on the target node features and / or the target edge features; Training the target multi-head graph attention network layer to perform a target update operation on the target node features and / or the target edge features; And training the target long short-term network layer to output target information to construct a target plastic failure fracture model according to the updated target node features and / or target edge features, where the above-mentioned target information includes: target displacement information, target stress information, and / or target strain information.

[0007] In some feasible embodiments, the above method further includes: Training the target multi-head graph attention network layer to determine the corresponding target attention weights based on the following formula: ; Among them, is the attention weight of node to its neighbor nodes ; is the unnormalized attention score between node and node ; is the result of exponentiating the unnormalized attention score; is the sum of the exponents of the attention scores of all neighbor nodes of node ; is the set of neighbors of node ; Among them, is determined based on the following formula: ; Among them, is the target learning vector; is the parameter vector corresponding to the th attention head; is the weight matrix corresponding to the th attention head; is the feature vector corresponding to the target node ; is the feature vector corresponding to the target node ; is the target activation function.

[0008] In some feasible embodiments, the above method further includes: Training the target multi-head graph attention network layer and performing a target update operation based on the following formula: ; Among them, is the updated representation vector of the target node at the th layer; is the activation function; is the attention weight of the target node to its neighbor node , which is used to represent the influence degree of the neighbor node on the target node ; is the weight matrix for feature mapping; is the representation vector of the neighbor node at the th layer; ; Among them, is the target node The updated representation vector of the first target layer; Used to represent the concatenation operation on the results of For the th attention head corresponding to the target node The attention weights for the neighbor nodes ; For the weight matrix corresponding to the th attention head; Is the representation vector of the neighbor node ; Wherein, Is the updated representation vector of the target node in the second target layer; Used to represent the average operation on For the th attention head corresponding to the target node The attention weights for the neighbor nodes ; For the weight matrix corresponding to the th attention head; Is the representation vector of the neighbor node Wherein, the first target layer is greater than 0 and less than the second target layer.

[0009] In some feasible embodiments, the above method further includes: Training the target prediction model based on the following target loss function: ; Wherein, Is the target mean absolute error; Is the number of samples; Is the true value corresponding to the target node ; Is the predicted value corresponding to the target node ;

[0010] In some feasible embodiments, the above generating the target plastic failure fracture prediction result according to the target prediction model includes: Generating a target failure element prediction result and / or a target fracture probability prediction result according to the target plastic failure fracture model.

[0011] In some feasible embodiments, the above method further includes: Determining the equivalent plastic strain corresponding to the target element according to the following formula: ; Among them, is the equivalent plastic strain, dimensionless; plastic strain rate tensor; ; Among them, is the equivalent plastic strain corresponding to the moment, dimensionless; is the equivalent plastic strain corresponding to the moment; is the increment of the plastic strain tensor, used to represent the change in plastic strain corresponding to the time step ; i and j are the direction indexes of the spatial coordinates; When it is determined that the equivalent plastic strain is greater than or equal to the preset failure strain, the target element is determined as the target failure element.

[0012] In some feasible embodiments, the above method further includes: Determine the target fracture probability prediction result according to the following formula: ; Among them, is the fracture probability of the target edge; is the first training parameter; is the second training parameter; is the activation function; target node feature vector.

[0013] In some feasible embodiments, the above method further includes: Optimize the target fracture probability prediction result according to the following formula: ; Among them, is the target optimization function; is the number of samples; is the target node corresponding true label; is the fracture probability predicted by the target prediction model.

[0014] In the second aspect of the present application, a plastic failure fracture prediction system is provided, including: A training unit for training a target prediction model according to a target feature dataset; A generating unit for generating a target plastic failure fracture prediction result according to the target prediction model; Among them, the target feature dataset includes: target node features and target edge features; The target prediction model includes: a target graph neural network layer, a target multi-head graph attention network layer, and a target long short-term network layer.

[0015] An embodiment of the present application provides a plastic failure fracture prediction method and system. The method includes: training a target prediction model according to a target feature dataset; generating a target plastic failure fracture prediction result according to the target prediction model. The target feature dataset includes: target node features and target edge features. The target prediction model includes: a target graph neural network layer, a target multi-head graph attention network layer, and a target long short-term network layer. The present application can reduce the consumption of computing resources in the plastic failure fracture prediction process, simplify the calculation process, and improve the plastic failure fracture simulation and iteration efficiency.

[0016] It should be understood that the content described in the summary of the invention section is not intended to limit the key or important features of the embodiments of the present application, nor is it used to limit the scope of the present application. Other features of the present application will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Combined with the drawings and referring to the following detailed description, the above and other features, advantages, and aspects of the embodiments of the present application will become more obvious. In the drawings, the same or similar reference numerals represent the same or similar elements, where: Figure 1 is a flowchart of a plastic failure fracture prediction method provided according to an embodiment of the present application; Figure 2 is a structural diagram of a plastic failure fracture prediction system provided according to an embodiment of the present application; Figure 3 is a structural diagram of a terminal device or a server suitable for implementing the embodiments of the present application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0018] To make the objectives, technical solutions, and advantages of the embodiments of the present disclosure clearer, the technical solutions in the embodiments of the present disclosure will be clearly and completely described below with reference to the drawings in the embodiments of the present disclosure. Obviously, the described embodiments are some, but not all, of the embodiments of the present disclosure. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present disclosure without creative efforts shall fall within the protection scope of the present disclosure.

[0019] In addition, the term "and / or" in this article is only a description of the association relationship of associated objects, indicating that there can be three relationships. For example, A and / or B can represent: A exists alone, A and B exist simultaneously, and B exists alone. In addition, the character " / " in this article generally represents an "or" relationship between the associated objects before and after.

[0020] Plastic metal fracture failure refers to the fracture failure that occurs after a metal material undergoes significant plastic deformation during the stress process. Among them, the plastic metal fracture failure is different from the brittle material fracture. Usually, large plastic deformation will occur before the plastic metal fractures, forming characteristics such as necking, micropore aggregation, and shear bands.

[0021] The plastic failure fracture process involves complex micro-damage evolution and non-linear large deformation. The research on plastic failure fracture is crucial in the fields of safety assessment of engineering structures, aerospace, and automobile manufacturing.

[0022] In related technologies, there are problems such as large consumption of computing resources, complex calculation process, and long simulation and iteration time in plastic failure fracture prediction.

[0023] In the first aspect of the present application, a plastic failure fracture prediction method is provided. Figure 1 It is a schematic flow chart of a plastic failure fracture prediction method 100 provided by an embodiment of the present application, as Figure 1 shown, the method 100 includes: Step S1; training a target prediction model according to a target feature data set.

[0024] It should be noted that the above target feature data set can be generated based on a large amount of simulation data of a finite element tool.

[0025] Exemplarily, the above target feature data set can be generated according to the simulation data of a target three-dimensional entity model under different working conditions.

[0026] Specifically, within the target parameter space, based on the Latin hypercube sampling method, perform target modeling operations and mesh generation operations to generate a target three-dimensional entity model. For different working conditions, adjust the target size parameters, target load parameters, target failure strain parameters, target material parameters, etc. of the target three-dimensional entity model, set corresponding target boundary conditions, and use the explicit dynamics algorithm for simulation calculation and solution to obtain the target feature data set. Among them, the above target material parameters may include: elastic modulus, Poisson's ratio, etc.

[0027] It should be noted that the above target feature data set may include: target node features and target edge features.

[0028] Exemplarily, the above target node features may include: node initial coordinate information, target value information, size parameters, load parameters corresponding to each node, and / or failure strain parameters, etc. Among them, the above target value information may include: displacement change information corresponding to each node in each frame, Von Mises stress field information, and / or equivalent plastic strain field time history data information.

[0029] Specifically, the above-mentioned target node features can be obtained by performing a target extraction operation on the data of the target three-dimensional entity model according to the input structure required by the network model.

[0030] Exemplarily, the above-mentioned target edge features may include: edge connection information, edge feature information, edge length information, edge vector information, and / or edge target value information, etc. Among them, the above-mentioned edge target value information may include: edge failure and fracture information.

[0031] Specifically, the above-mentioned target edge features can be obtained by performing a duplicate removal operation on the target node features for the data of the target three-dimensional entity model.

[0032] It should be noted that before performing step S1: training the target prediction model according to the target feature dataset, a target preprocessing operation can be performed on the above-mentioned target feature dataset. Among them, the above-mentioned target preprocessing operation may include: balancing processing, random shuffling processing, and / or normalization processing, etc.

[0033] Specifically, the above-mentioned target feature dataset can be divided into non-overlapping training sets, validation sets, and test sets. Among them, the proportion of the above-mentioned training set can be 70%, the proportion of the above-mentioned validation set can be 20%, and the proportion of the above-mentioned test set can be 10%.

[0034] Specifically, the mean and standard deviation corresponding to each data in the target feature dataset can be calculated batch by batch, and then the standardized data can be generated according to the mean and standard deviation.

[0035] Specifically, the mean can be calculated based on the following formula: (1)； Among them, is the mean, is the number of samples, is the value of each data point.

[0036] Specifically, the standard deviation can be calculated based on the following formula: (2)； Among them, is the standard deviation, is the mean, is the number of samples, is the value of each data point.

[0037] Specifically, the standardized data can be generated based on the following formula: (3)； Among them, is the original data, is the mean value, is the standard deviation, is the data after standardization.

[0038] In some feasible embodiments, the above-mentioned target prediction model includes: a target graph neural network layer, a target multi-head graph attention network layer, and a target long short-term network layer.

[0039] In some feasible embodiments, training the target prediction model according to the target feature dataset includes: training the target graph neural network layer for performing a target feature projection operation on the target node features and / or the target edge features.

[0040] Exemplarily, the standardized target feature dataset can be input into the encoder corresponding to the target graph neural network layer to perform a target feature projection operation on the target node features and / or the target edge features, so that the target node features and / or the target edge features are projected into a larger latent space.

[0041] It should be noted that the target graph neural network layer can be configured with a first encoder and a second encoder. Among them, the first encoder is used to perform an encoding operation on the target node features; the second encoder is used to perform an encoding operation on the edge features. Among them, each layer can use GeLU as the activation function and finally perform layer normalization based on GraphNorm to improve the stability of the target graph neural network layer of the model.

[0042] Among them, the target latent graph corresponding to the target feature dataset encoded by the target graph neural network layer can be input into the target multi-head graph attention network layer to achieve message passing.

[0043] In some feasible embodiments, training the target prediction model according to the target feature dataset further includes: training the target multi-head graph attention network layer for performing a target update operation on the target node features and / or the target edge features.

[0044] It should be noted that the above-mentioned target multi-head graph attention network layer can be used to achieve information interaction between target nodes.

[0045] In some feasible embodiments, the above method further includes: training the target multi-head graph attention network layer to determine the corresponding target attention weights based on the following formula: (4)； Among them, is the node for the neighbor node the attention weight of; is the node and the unnormalized attention score between the nodes ; is the result of exponentiating the unnormalized attention score; is the sum of the exponents of the attention scores of all neighbor nodes of node ; is the neighbor set of node ; wherein, is determined based on the following formula: (5)； wherein, is the target learning vector; is the parameter vector corresponding to the th attention head; is the weight matrix corresponding to the th attention head; is the feature vector corresponding to the target node ; is the feature vector corresponding to the target node ; is the target activation function.

[0046] It should be noted that the above belongs to the first real number matrix ; the above belongs to the second real number matrix .

[0047] Thus, the above method can accurately determine the attention weights independently calculated by each attention head based on the above formulas (4) to (5), which is beneficial to improving the expression ability of the target prediction model, improving the parallelism of the target prediction model, refining the information processing ability of the target prediction model, and improving the robustness of the target prediction model.

[0048] In some feasible implementation manners, the above method further includes: training the target multi-head graph attention network layer and performing a target update operation based on the following formula: (6)； wherein, is the updated representation vector of the target node at the th layer; is the activation function; is the attention weight of the target node for the neighbor node , which is used to represent the neighbor node for the target node Degree of influence; is the weight matrix for feature mapping; is the neighbor node at the layer representation vector.

[0049] It should be noted that formula (6) is used to perform the target update operation on each target node feature.

[0050] (7)； Among them, is the target node updated representation vector at the first target layer; is used to represent the concatenation operation on the results of different attention heads; is the th attention head corresponding target node attention weight for neighbor node ; is the th attention head corresponding weight matrix; is the neighbor node representation vector.

[0051] It should be noted that formula (7) is used to determine the concatenation value of each head corresponding to the intermediate layer.

[0052] (8)； Among them, is the target node updated representation vector at the second target layer; is used to represent the averaging operation on different attention heads; is the th attention head corresponding target node attention weight for neighbor node ; is the th attention head corresponding weight matrix; is the neighbor node representation vector; among them, the first target layer is greater than 0 and less than the second target layer.

[0053] It should be noted that formula (8) is used to determine the mean value of each head corresponding to the last layer.

[0054] Thus, based on formulas (6) - (8), the above method is conducive to performing target update operations on target node features and target edge features, thereby forming a new hidden state.

[0055] It should be noted that the above steps can be repeated rounds, that is, after Message Steps are performed, it is passed to the target long short - term network layer for temporal modeling.

[0056] In some feasible embodiments, training the target prediction model according to the target feature dataset further includes: training the target long short - term network layer, which is used to output target information based on the updated target node features and / or target edge features to construct a target plastic failure fracture model.

[0057] It should be noted that the target node features and / or target edge features, which are also the new hidden state after being processed by the target multi - head graph attention network layer, after being input into the target long short - term network layer, the target long short - term network layer can, based on the gating mechanism, retain long - term dependence information and suppress irrelevant short - term noise.

[0058] Among them, the target long short - term network layer can input the target node features and / or target edge node features processed by the target long short - term network layer into the node decoder to output target information, where the above target information includes: target displacement information, target stress information, and / or target strain information.

[0059] Among them, the above decoder structure is similar to the encoder, and the above decoder structure can add independent processing of each component for each channel to output the above target information for the final temporal physical field to construct a target plastic failure fracture model.

[0060] Thus, by configuring the target graph neural network layer, the target multi - head graph attention network layer, and the target long short - term network layer in the above target prediction model of the present application, it is conducive to reducing resource consumption in the plastic failure fracture simulation process to improve the simulation speed, simplifying the calculation process to reduce the calculation cost, improving the plastic failure fracture prediction accuracy. By training the target graph neural network layer, it is conducive to improving the execution accuracy and execution efficiency of the target feature projection operation corresponding to the target node features and / or target edge features. By training the target multi - head graph attention network layer, it is conducive to improving the execution accuracy and execution efficiency of the target update operation corresponding to the target node features and / or target edge features. By the target long short - term network layer, it is conducive to improving the output accuracy and output efficiency of the target information, thereby improving the construction accuracy and construction efficiency of the target plastic failure fracture model, and further improving the prediction accuracy and prediction efficiency of the target model.

[0061] In some feasible embodiments, the above method further includes: Train a target prediction model based on the following objective loss function: (9)； where is the target mean absolute error; is the number of samples; is the target node corresponding true value; is the target node corresponding predicted value.

[0062] It should be noted that the above target prediction model can be trained based on the objective loss function shown in formula (9), and the convergence of the target graph neural network layer can be judged by the validation set to further improve the training accuracy of the target prediction model.

[0063] Step S2; Generate a target plastic failure fracture prediction result according to the target prediction model.

[0064] Exemplarily, after completing the training operation of the target prediction model, the information corresponding to the metal material to be predicted can be input into the target prediction model to generate a target plastic failure fracture prediction result. Among them, the above target plastic failure fracture prediction result can include: a target failure element prediction result, a target fracture probability prediction result, and / or a target fracture mode prediction result.

[0065] In some feasible embodiments, the above generating a target plastic failure fracture prediction result according to the target prediction model includes: Generate a target failure element prediction result and / or a target fracture probability prediction result according to the target plastic failure fracture model.

[0066] It should be noted that the above target plastic failure fracture model can include: a target failure element prediction model and / or a target fracture probability prediction model.

[0067] In some feasible embodiments, the above method further includes: Determine the equivalent plastic strain corresponding to the target element according to the following formula: (10)； where is the equivalent plastic strain, dimensionless; is the plastic strain rate tensor; (11)； where is the equivalent plastic strain corresponding to the time, dimensionless; is The equivalent plastic strain corresponding to the moment; is the increment of the plastic strain tensor, used to represent the time step The corresponding change in plastic strain; i and j is the direction index of the spatial coordinate; When it is determined that the equivalent plastic strain is greater than or equal to the preset failure strain it is determined that the target element is the target failure element.

[0068] It should be noted that the above preset failure strain can correspond to the maximum equivalent plastic strain, that is, the material critical value. Among them, the above maximum equivalent plastic strain, that is, the material critical value, can be calibrated through experiments.

[0069] Exemplarily, based on finite element simulation, the failure strain parameter can be input as the critical value in the material parameters. Based on the above formula (10), when the equivalent plastic strain at the integration point of the target element reaches the material critical value, and / or, based on the above formula (11), when it is determined that the equivalent plastic strain at all integration points within the target element exceeds the above maximum equivalent plastic strain, it can be determined that the target element is the target failure element, and control is performed to delete the target failure element.

[0070] Thus, the above method can accurately determine the target failure element based on formulas (10) to (11), thereby accurately generating the prediction result of the target failure element according to the target plastic failure fracture model, improving the generation accuracy and generation efficiency of the prediction result of the target failure element, and further improving the prediction accuracy and prediction efficiency of plastic failure fracture.

[0071] In some feasible implementation manners, the above method further includes: Determine the target fracture probability prediction result according to the following formula: (12)； where is the fracture probability of the target edge; is the first training parameter; is the second training parameter; is the activation function; The target node feature vector.

[0072] It should be noted that the edge decoder can use a multi-layer perceptron as a classifier to calculate the fracture probability of the target edge based on the above formula (12) to generate the target fracture probability prediction result.

[0073] Thus, the above method can accurately determine the fracture probability of the target edge based on formula (12) to generate the target fracture probability prediction result, which is beneficial to improving the generation accuracy and efficiency of the target fracture probability prediction result.

[0074] In some feasible embodiments, the above method further includes: Optimizing the target fracture probability prediction result according to the following formula: (13)； wherein, is the target optimization function; is the number of samples; is the target node corresponding true label; is the fracture probability predicted by the target prediction model.

[0075] It should be noted that optimizing the target fracture probability prediction result based on the above formula (13) is beneficial to realizing iterative optimization of the target fracture probability prediction result and further improving the generation accuracy and efficiency of the target fracture probability prediction result.

[0076] Based on this, the embodiments of the present application provide a plastic failure fracture prediction method and system. The method includes: training a target prediction model according to a target feature dataset; generating a target plastic failure fracture prediction result according to the target prediction model; wherein the target feature dataset includes: target node features and target edge features; the above target prediction model includes: a target graph neural network layer, a target multi-head graph attention network layer, and a target long short-term network layer. The present application can achieve the following beneficial technical effects: (1) Combining a graph neural network and a long short-term network for predicting the physical field and fracture part of fracture failure, and using a multi-head graph attention network for information transmission is beneficial to improving the prediction accuracy, prediction efficiency, and generalization ability of the present application; (2) The target feature dataset includes: target node features and target edge features, and can accurately generate the target plastic failure fracture prediction result without constructing a complex dynamic time series graph based on a graph neural network; (3) It can achieve the combination of target edge fracture prediction and physical constraints, and can accurately generate the target plastic failure fracture prediction result without data-driven even in the case of target unit failure, which is beneficial to improving the robustness of the method of the present application.

[0077] It should be noted that for the foregoing method embodiments, for the sake of simple description, they are all expressed as a series of action combinations. However, those skilled in the art should know that this application is not limited by the described action sequence, because according to this application, certain steps can be performed in other sequences or simultaneously. Secondly, those skilled in the art should also know that the embodiments described in the specification are all optional embodiments, and the actions and modules involved are not necessarily essential to this application.

[0078] The above is the introduction of the method embodiments. The following further illustrates the solution of this application through system embodiments.

[0079] Figure 2 Fig. shows a structural schematic diagram of a plastic failure fracture prediction system 200 proposed by an embodiment of this application, as Figure 2 shown, the system 200 includes: a training unit 210 and a generation unit 220.

[0080] The training unit 210 is used to train a target prediction model according to a target feature dataset; The generation unit 220 is used to generate a target plastic failure fracture prediction result according to the target prediction model; Among them, the target feature dataset includes: target node features and target edge features; The target prediction model includes: a target graph neural network layer, a target multi-head graph attention network layer, and a target long short-term network layer.

[0081] Those skilled in the art can clearly understand that for the convenience and brevity of description, the specific working processes of the described modules can refer to the corresponding processes in the foregoing method embodiments, and will not be elaborated herein.

[0082] Figure 3 Fig. shows a structural schematic diagram of a terminal device or server suitable for implementing the embodiments of this application.

[0083] As Figure 3 shown, the terminal device or server includes a central processing unit (CPU) 301, which can perform various appropriate actions and processes according to the program stored in the read-only memory (ROM) 302 or the program loaded from the storage section 308 into the random access memory (RAM) 303. In the RAM 303, various programs and data required for the operation of the terminal device or server are also stored. The CPU 301, ROM 302, and RAM 303 are connected to each other through a bus 304. The input / output (I / O) interface 305 is also connected to the bus 304.

[0084] The following components are connected to the I / O interface 305: an input section 306 including a keyboard, a mouse, etc.; an output section 307 including a cathode ray tube (CRT), a liquid crystal display (LCD), etc. and a speaker, etc.; a storage section 308 including a hard disk, etc.; and a communication section 309 including a network interface card such as a LAN card, a modem, etc. The communication section 309 performs communication processing via a network such as the Internet. A drive 310 is also connected to the I / O interface 305 as required. A removable medium 311 such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc. is mounted on the drive 310 as required so that a computer program read therefrom is installed into the storage section 308 as required.

[0085] Specifically, according to an embodiment of the present application, the above method flow steps can be implemented as a computer software program. For example, an embodiment of the present application includes a computer program product including a computer program carried on a machine-readable medium, the computer program including program codes for performing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from a network via the communication section 309, and / or installed from the removable medium 311. When the computer program is executed by a central processing unit (CPU) 301, the above functions defined in the system of the present application are executed.

[0086] It should be noted that the computer-readable medium shown in this application can be a computer-readable signal medium, a computer-readable storage medium, or any combination of the two. A computer-readable storage medium can be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination of the above. More specific examples of a computer-readable storage medium can include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In this application, a computer-readable storage medium can be any tangible medium that contains or stores a program, and this program can be used by or in conjunction with an instruction execution system, apparatus, or device. And in this application, a computer-readable signal medium can include a data signal propagated in a baseband or as part of a carrier wave, which carries computer-readable program code. Such a propagated data signal can take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. A computer-readable signal medium can also be any computer-readable medium other than a computer-readable storage medium, and this computer-readable medium can send, propagate, or transmit a program for use by or in conjunction with an instruction execution system, apparatus, or device. The program code contained on a computer-readable medium can be transmitted using any appropriate medium, including but not limited to: wireless, wire, optical cable, RF, etc., or any suitable combination of the above.

[0087] The flowcharts and block diagrams in the accompanying drawings illustrate the possible architectures, functions, and operations of systems, methods, and computer program products according to various embodiments of this application. In this regard, each block in a flowchart or block diagram can represent a module, a program segment, or a part of code, and the aforementioned module, program segment, or part of code contains one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions marked in the blocks can occur in a different order than that marked in the accompanying drawings. For example, two consecutive blocks shown can actually be executed substantially in parallel, and they can sometimes be executed in the reverse order, depending on the functions involved. It should also be noted that each block in the block diagram and / or flowchart, as well as the combination of blocks in the block diagram and / or flowchart, can be implemented by a dedicated hardware-based system for performing the specified functions or operations, or can be implemented by a combination of dedicated hardware and computer instructions.

[0088] The units or modules involved in the embodiments described in this application can be implemented in software or in hardware. The described units or modules can also be provided in a processor. Among them, the names of these units or modules do not, in some cases, constitute a limitation on the units or modules themselves.

[0089] As another aspect, the present application also provides a computer-readable storage medium. The computer-readable storage medium can be included in the electronic device described in the foregoing embodiments; it can also exist alone without being assembled into the electronic device. The foregoing computer-readable storage medium stores one or more programs, and when the foregoing programs are executed by one or more processors, they implement the methods described in this application.

[0090] The above description is only a preferred embodiment of the present application and an explanation of the applied technical principles. Those skilled in the art should understand that the scope of the application involved in this application is not limited to the technical solutions formed by the specific combination of the above technical features, and should also cover other technical solutions formed by any combination of the above technical features or their equivalent features without departing from the foregoing inventive concept. For example, the technical solutions formed by mutually replacing the above features with other technical features (but not limited to) having similar functions in this application.

Claims

1. A method for predicting plastic failure fracture, characterized in that, Including: Training a target prediction model according to a target feature dataset; Generating a target plastic failure fracture prediction result according to the target prediction model; Wherein, the target feature dataset includes: target node features and target edge features; The target prediction model includes: a target graph neural network layer, a target multi-head graph attention network layer, and a target long short-term network layer.

2. The method for predicting plastic failure fracture according to claim 1, characterized in that, The training of the target prediction model according to the target feature dataset includes: Training the target graph neural network layer to perform a target feature projection operation on the target node features, and / or, the target edge features; Training the target multi-head graph attention network layer to perform a target update operation on the target node features, and / or, the target edge features; And training the target long short-term network layer to output target information according to the updated target node features and / or target edge features to construct a target plastic failure fracture model, wherein the target information includes: target displacement information, target stress information, and / or target strain information.

3. The method for predicting plastic failure fracture according to claim 2, characterized in that, Also including: Training the target multi-head graph attention network layer to determine corresponding target attention weights based on the following formula: ； Among them, is the node 's attention weight for neighbor nodes ; is the unnormalized attention score between node and node ; is the result of exponentiating the unnormalized attention score; is the sum of the exponents of the attention scores of all neighbor nodes of node ; is the neighbor set of node ; Among them, is determined based on the following formula: ； Among them, is the target learning vector; is the parameter vector corresponding to the -th attention head; is the weight matrix corresponding to the -th attention head; is the feature vector corresponding to the target node ; is the feature vector corresponding to the target node ; is the target activation function.

4. The method for predicting plastic failure fracture according to claim 3, characterized in that, Also including: Training the target multi-head graph attention network layer to perform the target update operation based on the following formula: ； Among them, is the target node at the updated representation vector of the layer; is the activation function; is the target node for the neighbor node attention weight, used to represent the neighbor node to the target node influence degree; is the weight matrix for feature mapping; is the neighbor node at the representation vector of the layer; ； Among them, is the target node The updated representation vector in the first target layer; Used to represent the Concatenation operation on the results of different attention heads; is the Target node corresponding to the th attention head Attention weights for neighbor nodes ; is the Weight matrix corresponding to the th attention head; is the representation vector of the neighbor node ; ； Among them, is the target node The updated representation vector in the second target layer; Used to represent the Average operation on different attention heads; is the Target node corresponding to the attention head Attention weights for neighbor nodes ; is the Weight matrix corresponding to the attention head; is the representation vector of the neighbor node ; Wherein, the first target layer is greater than 0 and less than the second target layer.

5. The method for predicting plastic failure fracture according to claim 4, characterized in that, Also including: Training the target prediction model based on the following target loss function: ； Among them, is the target mean absolute error; is the number of samples; is the true value corresponding to the target node ; is the predicted value corresponding to the target node .

6. The method for predicting plastic failure fracture according to any one of claims 2 to 5, characterized in that, The generating of the target plastic failure fracture prediction result according to the target prediction model includes: Generating a target failure unit prediction result and / or a target fracture probability prediction result according to the target plastic failure fracture model.

7. The plastic failure fracture prediction method according to claim 6, wherein, Also including: Determining the equivalent plastic strain corresponding to a target unit according to the following formula: ； Among them, is the equivalent plastic strain, dimensionless; plastic strain rate tensor; ； Among them, is the equivalent plastic strain corresponding to the moment, dimensionless; is the equivalent plastic strain corresponding to the moment; is the increment of the plastic strain tensor, used to represent the change of the plastic strain corresponding to the time step ; i and j are the direction indices of the spatial coordinates; When it is determined that the equivalent plastic strain is greater than or equal to a preset failure strain, determining that the target unit is the target failure unit.

8. The plastic failure fracture prediction method according to claim 7, wherein, Also including: Determining the target fracture probability prediction result according to the following formula: ； Among them, is the fracture probability of the target edge; is the first training parameter; is the second training parameter; is the activation function; Target node feature vector.

9. The plastic failure fracture prediction method according to claim 8, wherein, Also including: Optimizing the target fracture probability prediction result according to the following formula: ； Among them, is the target optimization function; is the number of samples; is the target node corresponding true label; is the fracture probability predicted by the target prediction model.

10. A plastic failure fracture prediction system, wherein, Including: A training unit for training a target prediction model according to a target feature dataset; A generating unit for generating a target plastic failure fracture prediction result according to the target prediction model; Wherein, the target feature dataset includes: target node features and target edge features; The target prediction model includes: a target graph neural network layer, a target multi-head graph attention network layer, and a target long short-term network layer.

Images