A method, device, medium and product for generating a three-dimensional microstructure image of a titanium alloy

Abstract

The application discloses a titanium alloy three-dimensional microstructure image generation method and device, medium and product, relates to the titanium alloy three-dimensional microstructure reconstruction field, and the method comprises the following steps: constructing a titanium alloy three-dimensional grain sample set; training a three-dimensional microstructure image generation model according to the sample set, and obtaining a trained three-dimensional microstructure image generation model; the three-dimensional microstructure image generation model comprises a feature vector extraction layer, a plane parameter extraction layer, a plane extraction layer, a plane grouping layer, a microstructure generation layer and a microstructure screening layer; and the trained three-dimensional microstructure image generation model is used for generating a three-dimensional microstructure image of a titanium alloy test piece. The application can intuitively and accurately reproduce similar microstructure images, and realizes intelligent generation of virtual microstructure.

Description

Technical Field

[0001] This application relates to the field of three-dimensional microstructure reconstruction of titanium alloys, and in particular to a method, device, medium and product for generating three-dimensional microstructure images of titanium alloys. Background Technology

[0002] Titanium alloys possess excellent corrosion resistance, lightweight yet high strength, and high-temperature stability, making them widely used in military aircraft, aero engines, spacecraft fuel tanks, and missile components. For example, the composition of titanium alloys used in turbine engine components is diverse, and the type of titanium alloy used can vary significantly depending on the location and operational requirements of a specific turbine section. Understanding grain structure and properties is crucial for in-depth research into material properties. Therefore, analyzing and studying material grains can better reveal the material's performance characteristics and variation patterns, providing important references and bases for material design and applications.

[0003] The microstructure of materials is inextricably linked to their macroscopic properties, and accurately obtaining the microstructure is crucial for achieving full lifecycle quality control of critical components. However, existing experimental characterization techniques for microstructure can damage critical components, have long testing cycles, and are costly. Furthermore, microstructure simulation techniques require significant computational resources and the models are not accurate enough.

[0004] Therefore, based on the above problems, there is an urgent need to provide a method for generating three-dimensional microstructure images of titanium alloys, so as to solve the problems of scarce data samples and the cumbersome, time-consuming and costly process of obtaining three-dimensional microstructure images of alloys in the existing technology, and thus be able to intuitively and accurately reproduce similar microstructure images and realize the intelligent generation of virtual microstructures. Summary of the Invention

[0005] The purpose of this application is to provide a method, device, medium, and product for generating three-dimensional microstructure images of titanium alloys, which can intuitively and accurately reproduce similar microstructure images and realize the intelligent generation of virtual microstructures.

[0006] To achieve the above objectives, this application provides the following solution:

[0007] In a first aspect, this application provides a method for generating a three-dimensional microstructure image of a titanium alloy, the method comprising:

[0008] A sample set of three-dimensional grains of titanium alloy is constructed; each sample in the sample set includes: a three-dimensional microstructure image of titanium alloy and corresponding geometric structural feature information;

[0009] A three-dimensional microstructure image generation model is trained based on a sample set to obtain a trained three-dimensional microstructure image generation model. The three-dimensional microstructure image generation model includes: a feature vector extraction layer, a plane parameter extraction layer, a plane extraction layer, a plane grouping layer, a microstructure generation layer, and a microstructure filtering layer. The feature vector extraction layer extracts the feature vector for each sample. The plane parameter extraction layer generates a plane parameter matrix based on the feature vector and a multilayer perceptron. The plane extraction layer determines the signed distance between each point and each plane based on the coordinates of points in the random sampling space, using the plane parameter matrix. The plane grouping layer aggregates planes into convex polyhedra based on the signed distance between each point and each plane and the binary matrix, using max pooling to obtain a convex set. The microstructure generation layer assembles the convex set into a three-dimensional shape. The microstructure filtering layer filters based on the three-dimensional shape using 3D moment invariants to obtain a three-dimensional microstructure image.

[0010] A trained three-dimensional microstructure image generation model was used to generate three-dimensional microstructure images of titanium alloy specimens.

[0011] Optionally, the geometric structural features include: grain size, crystal orientation, morphological parameters, and the number of adjacent grains.

[0012] Optionally, the sample set for constructing three-dimensional grains of titanium alloy further includes:

[0013] Perform format conversion on the sample set.

[0014] Alternatively, the formula D = xP can be used. T Determine the signed distance D between point x and the plane;

[0015] Where P is a plane parameter, f is the eigenvector. For a multilayer perceptron, the superscript T stands for transpose, and ω represents the parameters of each layer of the MLP.

[0016] Optionally, using the formula Aggregate planes into convex polyhedra;

[0017] in, Let D be the convex polyhedron corresponding to point x. i Let D represent the signed distance from point x to the i-th plane. i If x < 0, then point x lies in the plane; otherwise, it lies outside the plane. ij is a binary number in a binary matrix, used to characterize whether the i-th plane belongs to the j-th primitive, and max is the function to find the maximum value.

[0018] Optionally, using the formula Assemble convex sets into a three-dimensional shape S. * (x);

[0019] in, Let x be the position of point x relative to the j-th primitive, and min be the function that takes the minimum value.

[0020] Optionally, utilize Determine the 3D moment invariants;

[0021] Where θ1, θ2, and θ3 are nonnormalized moment invariants, V is the volume of the object, and Ω1, Ω2, and Ω3 are normalized moment invariants.

[0022] Secondly, this application provides a computer device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the method for generating images of three-dimensional microstructures of titanium alloys.

[0023] Thirdly, this application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method for generating images of the three-dimensional microstructure of titanium alloys.

[0024] Fourthly, this application provides a computer program product, including a computer program that, when executed by a processor, implements the method for generating images of the three-dimensional microstructure of titanium alloys.

[0025] According to the specific embodiments provided in this application, this application has the following technical effects:

[0026] This application provides a method, device, medium, and product for generating three-dimensional microstructure images of titanium alloys. Through a feature vector extraction layer, a plane parameter extraction layer, a plane extraction layer, a plane grouping layer, a microstructure generation layer, and a microstructure screening layer in the three-dimensional microstructure image generation model, it achieves the generation of corresponding virtual three-dimensional microstructure images based on real-time titanium alloy three-dimensional microstructure images. This solves the problems of scarce data samples and the cumbersome, time-consuming, and costly process of obtaining the internal three-dimensional microstructure of alloys in existing technologies. By employing the 3D moment invariant principle to screen virtual microstructure images, the microstructure images of titanium alloy specimens are obtained. This allows for the intuitive and accurate reproduction of similar microstructure images, realizing the intelligent generation of virtual microstructures and bringing convenience to the study of material properties. Attached Figure Description

[0027] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0028] Figure 1 This is a schematic diagram of a method for generating a three-dimensional microstructure image of a titanium alloy according to an embodiment of this application;

[0029] Figure 2 This is a schematic diagram illustrating the principle of a three-dimensional microscopic tissue structure image generation model. Detailed Implementation

[0030] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0031] To make the above-mentioned objectives, features and advantages of this application more apparent and understandable, the application will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0032] In one exemplary embodiment, such as Figure 1 As shown, a method for generating three-dimensional microstructure images of titanium alloys is provided, the method comprising the following steps S101 to S103. Wherein:

[0033] S101, construct a sample set of three-dimensional grains of titanium alloy; each sample in the sample set includes: a three-dimensional microstructure image of titanium alloy and corresponding geometric structural feature information;

[0034] The geometric structural features include: grain size, crystal orientation, morphological parameters, and the number of adjacent grains.

[0035] Following S101 are:

[0036] Perform format conversion on the sample set.

[0037] The specific format conversion process is as follows:

[0038] First, the .obj format files of the sample set are converted to .mat files using MATLAB code. Then, the .mat files are further converted to .hdf5 format using Python code. Finally, the format-converted sample set is input into the feature vector extraction layer of the three-dimensional microstructure image generation model.

[0039] S102, Based on the sample set, train the three-dimensional microstructure image generation model to obtain the trained three-dimensional microstructure image generation model; such as Figure 2 As shown, the three-dimensional microstructure image generation model includes: a feature vector extraction layer, a plane parameter extraction layer (L0 layer), a plane extraction layer (L1 layer), a plane grouping layer (L2 layer), a microstructure generation layer (L3 layer), and a microstructure screening layer; the feature vector extraction layer is used to extract the feature vector of each sample; the plane parameter extraction layer is used to generate a plane parameter matrix P based on the feature vector and a multilayer perceptron (MLP). p×4 The plane extraction layer is used to determine the signed distance between each point and each plane based on the coordinates of points in the random sampling space, according to the matrix of plane parameters; the plane grouping layer is used to determine the signed distance between each point and each plane and the binary matrix T. p×c Based on the max pooling operation, planes are aggregated into convex polyhedra to obtain convex sets; the microstructure generation layer is used to assemble three-dimensional shapes according to the convex sets; the microstructure filtering layer is used to filter according to the three-dimensional shapes based on 3D moment invariants to obtain three-dimensional microstructure images.

[0040] As a specific implementation, the feature vector extraction layer is an SANet network. SANet effectively captures global contextual relationships, dynamically adjusts feature weights, avoids information loss, handles multi-scale information, and improves computational efficiency by processing local information. These characteristics enable SANet to possess powerful feature extraction capabilities in multiple tasks, especially in vision and sequence tasks.

[0041] The SANet network consists of two encoder layers and a fully connected output layer. Each encoder layer comprises a multi-head self-attention mechanism, LayerNorm, and a fully connected feedforward network. Furthermore, a shortcut structure from the residual network is also added to the SANet network.

[0042] Multi-head self-attention is a core component of the SANet network, allowing the model to collectively focus on information from different representation subspaces at different locations. The output H of the i-th attention head... k,i Represented as:

[0043]

[0044] in, and Let Q represent the linear projection of the i-th attention head. k K k and V k These are the k-th input sample FSS. k The query, key, and value are obtained through FSS. k The sum of input embedding and positional encoding, i.e., EFSS k The projection is obtained from Q. k =(EFSS) k W Q K k =(EFSS) k W K V k =(EFSS) k W V W Q W K and W V The parameter matrix representing the linear protrusion has a size d. model ×d Q d model ×d K and d model ×d V .

[0045] A multi-head self-attention mechanism consisting of h attention heads is represented as follows:

[0046] MH k =MultiHead(Q k ,K k V k =Concat(H) k,1 H k,2 ,…,H k,h W O ;

[0047] Among them, W O This represents the linear projection over all attention heads connected.

[0048] After processing by the SANet network, the fully extracted feature vectors are obtained.

[0049] In another embodiment, the matrix P of the plane parameters p×4 Define the implicit equations for p planes as follows:

[0050] ax + by + cz + d = 0;

[0051] The implicit equation is evaluated on a set of n point coordinates in the L1 layer, where the coordinates of the n points are x. n×4Multiplying by P gives the signed distance between point x and the plane.

[0052] That is, for any point x = (x, y, z, 1), randomly sampling n points in space yields an n x 4 matrix, denoted as x. n×4 The coordinates x of the n sampled points n×4 Multiplying by P yields the signed distance D between point x and the plane. If point x lies within the i-th plane, the distance is negative; otherwise, it is positive. Points within the plane are filled with their corresponding ranges; otherwise, no processing is needed. Here, P is a plane parameter. In this embodiment, P... p×4 p is the number of planes. f is the eigenvector. For a multilayer perceptron, the superscript T is the transpose, and ω is {W1,b1,W2,b2…}, which is the weight matrix W and bias vector b of each layer of the MLP.

[0053] To group the hyperplane into geometric primitives, the input plane is aggregated to form a set of c convex primitives, using a binary matrix T. p×c T p×c This indicates whether there is a connection between each plane and each convex polygon, and is used as the parameter P of the connecting hyperplane. p×4 The key bridge to the final shape is the forced selective neuron feeding from the L1 layer to the next L2 layer, forming the convex part; a connection is 1, and no connection is 0; p represents p planes, and c represents c convex polygons;

[0054] Specifically, using formulas Aggregate planes into convex polyhedra;

[0055] in, Let D be the convex polyhedron corresponding to point x. i Let D represent the signed distance from point x to the i-th plane. i If x < 0, then point x lies in the plane; otherwise, it lies outside the plane. ij is a binary number in a binary matrix, used to characterize whether the i-th plane belongs to the j-th primitive, and max is the function to find the maximum value.

[0056] During training, the gradient flows through only one (maximum) plane. Therefore, to simplify training, we use a plane where max is replaced with sum:

[0057] Here, ReLU is the activation function.

[0058] Mini-pooling is used to group convex groups to create a possible non-convex output shape:

[0059]

[0060] in, Let x be the position of point x relative to the j-th primitive, and min be the function that takes the minimum value.

[0061] To avoid inefficient memory utilization during training and to promote learning, a (weighted) summation method is used to distribute the gradient to all convex sets, resulting in the implicit function of the final 3D model:

[0062]

[0063] Among them, W c×1 For the weight matrix, [·] [0,1] , represents the division of the interval, S + (x) is only a case of whether a point is inside the model. There exists a point that is outside any convex polyhedron but inside their combined model.

[0064] To filter out unsuitable 3D microstructure images, using Determine the 3D moment invariants;

[0065] Where θ1, θ2, θ3 are nonnormalized moment invariants, V is the volume of the object, and Ω1, Ω2, Ω3 are normalized moment invariants.

[0066] S103, using a trained three-dimensional microstructure image generation model to generate three-dimensional microstructure images of titanium alloy specimens.

[0067] This application acquires three-dimensional microstructure images of titanium and evaluates the validity of all acquired sample data. Then, it preprocesses the microstructure images to obtain a specific file format and calculates the corresponding structural features using SANet. Based on the feature vectors and MLP, it generates a matrix P of normalized planar parameters in layer L0. p×4 The plane parameter defines the implicit equation for p planes: ax + by + cz + d = 0; the implicit equation represents the coordinates of n points x in layer L1. n×4 The set is evaluated; a binary matrix T is used. p×c The process involves forcing selective neuron feeding from L1 to the next L2 layer to form convex portions; the L3 layer assembles these portions into a shape through summation or minimum pooling; the trained model reconstructs a virtual three-dimensional microstructure; and the acquired virtual three-dimensional microstructure images are then detected and filtered to remove those that do not meet the requirements. This application addresses the difficulty of acquiring existing three-dimensional microstructure images by introducing a deep learning model, enabling convenient and rapid acquisition of three-dimensional microstructure images.

[0068] In one exemplary embodiment, a computer device is provided, which may be a server or a terminal. The computer device includes a processor, memory, input / output (I / O) interfaces, and a communication interface. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface is connected to the system bus via the I / O interfaces. The processor of the computer device provides computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores an operating system, computer programs, and a database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The I / O interfaces of the computer device are used for exchanging information between the processor and external devices. The communication interface of the computer device is used for communication with external terminals via a network connection. When the computer program is executed by the processor, it implements a method for generating three-dimensional microstructure images of titanium alloys.

[0069] In one exemplary embodiment, a computer-readable storage medium is provided storing a computer program that, when executed by a processor, implements the steps in the above-described method embodiments.

[0070] In one exemplary embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in the above-described method embodiments.

[0071] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data must comply with relevant regulations.

[0072] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM).

[0073] The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.

[0074] In this application, all actions to acquire signals, information, or data are carried out in compliance with the relevant data protection laws and policies of the country where the location is situated, and with the authorization granted by the owner of the relevant device.

[0075] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0076] This document uses specific examples to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the methods and core ideas of this application. Furthermore, those skilled in the art will recognize that, based on the ideas of this application, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A method for generating three-dimensional microstructure images of titanium alloys, characterized in that, The method for generating three-dimensional microstructure images of titanium alloys includes: A sample set of three-dimensional grains of titanium alloy is constructed; each sample in the sample set includes: a three-dimensional microstructure image of titanium alloy and corresponding geometric structural feature information; A three-dimensional microstructure image generation model is trained based on a sample set to obtain a trained three-dimensional microstructure image generation model. The three-dimensional microstructure image generation model includes: a feature vector extraction layer, a plane parameter extraction layer, a plane extraction layer, a plane grouping layer, a microstructure generation layer, and a microstructure filtering layer. The feature vector extraction layer extracts the feature vector for each sample. The plane parameter extraction layer generates a plane parameter matrix based on the feature vector and a multilayer perceptron. The plane extraction layer determines the signed distance between each point and each plane based on the coordinates of points in the random sampling space, using the plane parameter matrix. The plane grouping layer aggregates planes into convex polyhedra based on the signed distance between each point and each plane and the binary matrix, using max pooling to obtain a convex set. The microstructure generation layer assembles the convex set into a three-dimensional shape. The microstructure filtering layer filters based on the three-dimensional shape using 3D moment invariants to obtain a three-dimensional microstructure image. A trained three-dimensional microstructure image generation model was used to generate three-dimensional microstructure images of titanium alloy specimens.

2. The method for generating three-dimensional microstructure images of titanium alloys according to claim 1, characterized in that, The geometric structural features include: grain size, crystal orientation, morphological parameters, and the number of adjacent grains.

3. The method for generating three-dimensional microstructure images of titanium alloys according to claim 1, characterized in that, The sample set for constructing three-dimensional grains of titanium alloys further includes: Perform format conversion on the sample set.

4. The method for generating a three-dimensional microstructure image of titanium alloy according to claim 1, characterized in that, Using the formula D = xP T Determine the signed distance D between point x and the plane; Where P is a plane parameter, f is the eigenvector. For a multilayer perceptron, the superscript T stands for transpose, and ω represents the parameters of each layer of the MLP.

5. The method for generating three-dimensional microstructure images of titanium alloys according to claim 1, characterized in that, Using formula Aggregate planes into convex polyhedra; in, Let D be the convex polyhedron corresponding to point x. i Let D represent the signed distance from point x to the i-th plane. i If x < 0, then point x lies in the plane; otherwise, it lies outside the plane. ij is a binary number in a binary matrix, used to characterize whether the i-th plane belongs to the j-th primitive, and max is the function to find the maximum value.

6. The method for generating a three-dimensional microstructure image of titanium alloy according to claim 1, characterized in that, Using formula Assemble convex sets into a three-dimensional shape S. * (x); in, Let x be the position of point x relative to the j-th primitive, and min be the function that takes the minimum value.

7. The method for generating a three-dimensional microstructure image of titanium alloy according to claim 1, characterized in that, use Determine the 3D moment invariants; Where θ1, θ2, and θ3 are nonnormalized moment invariants, V is the volume of the object, and Ω1, Ω2, and Ω3 are normalized moment invariants.

8. A computer device, comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that the processor executes the computer program to implement the method for generating images of three-dimensional microstructures of titanium alloys according to any one of claims 1-7.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by a processor, the computer program implements the method for generating three-dimensional microstructure images of titanium alloys as described in any one of claims 1-7.

10. A computer program product, comprising a computer program, characterized in that, When executed by a processor, the computer program implements the method for generating three-dimensional microstructure images of titanium alloys as described in any one of claims 1-7.

Images