Nickel-based superalloy microstructure image diffusion generation method, system, equipment, medium and product

Abstract

The invention discloses a nickel-based superalloy microstructure image diffusion generation method, system and equipment, a medium and a product, and relates to the field of superalloy microstructure reconstruction. The method comprises the following steps: gradually diffusing an original training image into a Gaussian noise-like image by adopting a forward diffusion process of a denoising diffusion generation network, and gradually denoising in a reverse diffusion process to obtain a nickel-based high-temperature alloy microstructure-like image; adopting a geometric structure feature loss penalty mechanism to obtain a trained denoising diffusion generation network based on the geometric feature difference value; and adopting a back diffusion process of the trained de-noising diffusion generation network to obtain a virtual microstructure image, and screening to obtain a final microstructure image. According to the method, the problems that data samples are scarce, and acquisition of the alloy internal microstructure is tedious, time-consuming, high in cost and the like are solved, similar microstructure images are visually represented, and intelligent generation of the virtual microstructure is achieved.

Description

Technical Field

[0001] The present application relates to the field of microstructure reconstruction of high-temperature alloys, and in particular to a method, system, equipment, medium and product for diffusion generation of microstructure images of nickel-based high-temperature alloys. Background Art

[0002] Nickel-based high-temperature alloys have excellent high-temperature strength, corrosion resistance and thermal stability. They are widely used in aerospace and new energy industries. They are the main raw materials for hot-end critical parts such as aircraft engine turbine disks and combustion chamber blades. Critical parts are in high temperature, high pressure and high speed environments for a long time. Fatigue will cause changes in the microstructure of the critical parts material, and gradually develop from tiny defects to macroscopic defects until failure damage occurs. Therefore, changes in the organizational structure are the root cause of damage and failure.

[0003] There is an inseparable connection between the microstructure of materials and their macroscopic properties. Accurately obtaining the microstructure of materials is the key to achieving quality control of critical parts throughout their life cycle. However, the existing experimental characterization technology of microstructures is harmful to critical parts and has a long detection cycle and high detection costs. In addition, the simulation technology of microstructures does not accurately predict the large model required for computing resources. Summary of the invention

[0004] The purpose of this application is to provide a method, system, equipment, medium and product for diffusion generation of microstructure image of nickel-based high-temperature alloy, which can solve the problems faced by the prior art, such as scarcity of data samples and cumbersome, time-consuming and costly acquisition of the internal microstructure of the alloy, and can intuitively reproduce similar microstructure images, bringing convenience to the study of material properties.

[0005] To achieve the above objectives, this application provides the following solutions:

[0006] In a first aspect, the present application provides a method for diffusion generation of a nickel-based high-temperature alloy microstructure image, comprising:

[0007] Acquire multiple original samples of nickel-based high-temperature alloys to form a sample set; each original sample of the nickel-based high-temperature alloy includes: an original training image and geometric structure feature information corresponding to the original training image; the original training image is a microstructure image of the nickel-based high-temperature alloy; the geometric structure feature information includes grain size, grain roundness, and major-minor axis ratio of the grain;

[0008] The original training image is gradually diffused into a Gaussian noise-like image by using a forward diffusion process of a denoising diffusion generation network;

[0009] The Gaussian noise-like image is gradually denoised by using the reverse diffusion process of the denoising diffusion generation network to obtain a microstructure image of the nickel-based high-temperature alloy;

[0010] Determine the geometric structure feature information of the nickel-based high-temperature alloy microstructure image, and determine the geometric feature difference based on the geometric structure feature information corresponding to the original training image and the geometric structure feature information of the nickel-based high-temperature alloy microstructure image;

[0011] Determining whether the geometric feature difference is zero;

[0012] When the geometric feature difference is not zero, the geometric feature difference is introduced into the denoising diffusion model as a geometric structure feature loss penalty term to obtain a trained denoising diffusion generation network;

[0013] When the geometric feature difference is zero, zero is introduced into the denoising diffusion model as a geometric structure feature loss penalty term to obtain a trained denoising diffusion generation network;

[0014] Acquire an original image of a nickel-based high-temperature alloy specimen; the original image of the nickel-based high-temperature alloy specimen is a microstructure image containing Gaussian noise;

[0015] Generate a virtual microstructure image based on the original image of the nickel-based high-temperature alloy specimen by using a reverse diffusion process of a trained denoising diffusion generation network;

[0016] The virtual microstructure image is screened using the grain boundary integrity principle to obtain the microstructure image of the nickel-based high-temperature alloy specimen.

[0017] Optionally, the forward diffusion process of the denoising diffusion generation network is used to gradually diffuse the original training image into a Gaussian noise-like image, specifically including:

[0018] The forward diffusion process of the denoising diffusion generation network is adopted to add Gaussian noise to the original training image according to a set time step to obtain the Gaussian noise-like image.

[0019] Optionally, the forward diffusion process of the denoising diffusion generation network is a Markov process.

[0020] Optionally, the Gaussian noise-like image is gradually denoised by a reverse diffusion process of a denoising diffusion generation network to obtain a microstructure image of a nickel-based high-temperature alloy, specifically comprising:

[0021] Inputting the Gaussian noise-like image and the set time step into a U-Net prediction network to obtain a noise-added noise;

[0022] The reverse diffusion process of the denoising diffusion generation network is adopted to remove the noise in the Gaussian noise-like image according to the set time step to obtain the microstructure image of the nickel-based high-temperature alloy.

[0023] Optionally, determining the geometric structure feature information of the nickel-based high-temperature alloy microstructure image specifically includes:

[0024] Determine the grain area in the nickel-based high-temperature alloy microstructure image by using a contour area calculation method, and determine the grain size of a single grain based on the grain area;

[0025] Determine the grain length in the nickel-based high-temperature alloy microstructure image by using a contour length calculation method, and determine the roundness of a single grain based on the grain area and the grain length;

[0026] The major axis and the minor axis of the grains in the microstructure image of the nickel-based high-temperature alloy are determined by using an ellipse focus coordinate calculation method, and the major-minor axis ratio of a single grain is determined based on the major axis and the minor axis of the grain;

[0027] Obtaining an average grain size, an average roundness, and an average major-minor axis ratio based on the grain size of a single grain, the roundness of a single grain, and the major-minor axis ratio of a single grain;

[0028] The average grain size, average roundness and average major-minor axis ratio are used as geometric structural feature information of the microstructure image of the nickel-based high-temperature alloy.

[0029] Optionally, the virtual microstructure image is screened using the grain boundary integrity principle to obtain the microstructure image of the nickel-based high-temperature alloy specimen, specifically comprising:

[0030] Binarizing the virtual microscopic tissue structure image to obtain a binary image;

[0031] By means of image traversal, each pixel in the binary image is traversed, and an 8-neighborhood detection method is used to detect whether there is a 0 value near each 0-value pixel in the binary image; the vicinity of each 0-value pixel refers to an area extending from the 0-value pixel as a starting point to a set value;

[0032] When there is no 0 value near each 0-value pixel, it indicates that there is a break in the grain boundary line, and the virtual microstructure image is deleted;

[0033] When there is no 0 value near each 0-value pixel, it indicates that the grain boundaries in the entire microstructure are intact, and the virtual microstructure image is retained, and the retained virtual microstructure image is used as the microstructure image of the nickel-based high-temperature alloy specimen.

[0034] In a second aspect, the present application provides a system for generating a diffusion image of a nickel-based high-temperature alloy microstructure, comprising:

[0035] A data acquisition module, used to acquire multiple original samples of nickel-based high-temperature alloys to form a sample set, and to acquire original images of nickel-based high-temperature alloy specimens; each original sample of the nickel-based high-temperature alloy includes: an original training image and geometric structure feature information corresponding to the original training image; the original training image is a microstructure image of the nickel-based high-temperature alloy; the geometric structure feature information includes grain size, grain roundness, and major-minor axis ratio of the grain; the original image of the nickel-based high-temperature alloy specimen is a microstructure image containing Gaussian noise;

[0036] A forward diffusion module, used for gradually diffusing the original training image into a Gaussian noise-like image by using a forward diffusion process of a denoising diffusion generation network;

[0037] A reverse diffusion module is used to gradually denoise the Gaussian noise image by using a reverse diffusion process of a denoising diffusion generation network to obtain a microstructure image of a nickel-based high-temperature alloy;

[0038] A geometric feature difference determination module, used to determine the geometric structure feature information of the nickel-based high-temperature alloy microstructure image, and determine the geometric feature difference based on the geometric structure feature information corresponding to the original training image and the geometric structure feature information of the nickel-based high-temperature alloy microstructure image;

[0039] A geometric structure feature loss penalty module, used to determine whether the geometric feature difference is zero; when the geometric feature difference is not zero, the geometric feature difference is introduced into the denoising diffusion model as a geometric structure feature loss penalty term to obtain a trained denoising diffusion generation network; when the geometric feature difference is zero, zero is introduced into the denoising diffusion model as a geometric structure feature loss penalty term to obtain a trained denoising diffusion generation network;

[0040] A microstructure generation module, for generating a virtual microstructure image based on an original image of the nickel-based high-temperature alloy specimen by using a reverse diffusion process of a trained denoising diffusion generation network;

[0041] The microstructure image generation module is used to screen the virtual microstructure image by adopting the grain boundary integrity principle to obtain the microstructure image of the nickel-based high-temperature alloy specimen.

[0042] In a third aspect, the present application provides a computer device, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that the processor executes the computer program to implement the above-mentioned method for generating diffusion images of the microstructure structure of a nickel-based high-temperature alloy.

[0043] In a fourth aspect, the present application provides a computer-readable storage medium having a computer program stored thereon, characterized in that when the computer program is executed by a processor, the above-mentioned method for generating diffusion images of microstructure structures of nickel-based high-temperature alloys is implemented.

[0044] In a fifth aspect, the present application provides a computer program product, including a computer program, characterized in that when the computer program is executed by a processor, the above-mentioned method for generating diffusion images of microstructure structures of nickel-based high-temperature alloys is implemented.

[0045] According to the specific embodiments provided in this application, this application discloses the following technical effects:

[0046] The present application provides a method, system, device, medium and product for diffusion generation of microstructure image of nickel-based high-temperature alloy, which generates virtual microstructure image based on the original image of nickel-based high-temperature alloy specimen by adopting the reverse diffusion process of trained denoising diffusion generation network, and can obtain the corresponding virtual microstructure image based on the original image returned in real time, so as to solve the problems faced by the prior art that data samples are scarce and it is cumbersome, time-consuming and costly to obtain the microstructure inside the alloy. The virtual microstructure image is screened by adopting the grain boundary integrity principle to obtain the microstructure image of nickel-based high-temperature alloy specimen, which can intuitively reproduce similar microstructure images, realize the intelligent generation of virtual microstructure, and bring convenience to the research of material properties. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0048] Figure 1 This is an application environment diagram of a method for diffusion generation of a microstructure image of a nickel-based high-temperature alloy in one embodiment of the present application;

[0049] Figure 2 A flow chart of a method for diffusion generation of a microstructure image of a nickel-based high-temperature alloy provided in one embodiment of the present application;

[0050] Figure 3 A flow chart of a method for generating and visualizing the diffusion of a microstructure of GH4169 provided in another embodiment of the present application;

[0051] Figure 4 A schematic diagram of a method for generating and visualizing the diffusion of microstructures of GH4169 is provided for another embodiment of the present application;

[0052] Figure 5 Another embodiment of the present application provides a microstructure diagram of GH4169 collected through an optical microscope after a specified solution aging treatment;

[0053] Figure 6 A flow chart for calculating the geometrical characteristics of a microstructure provided by another embodiment of the present application;

[0054] Figure 7 A schematic diagram of functional modules of a system for generating a diffusion image of a nickel-based high-temperature alloy microstructure provided by another embodiment of the present application;

[0055] Figure 8 A system structure diagram of a microstructure diffusion generation and two-dimensional visualization method for GH4169 provided in another embodiment of the present application;

[0056] Fig. 9 A schematic diagram of the structure of a computer device provided in one embodiment of the present application. DETAILED DESCRIPTION

[0057] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0058] In order to make the above-mentioned objects, features and advantages of the present application more obvious and easy to understand, the present application is further described in detail below in conjunction with the accompanying drawings and specific implementation methods.

[0059] The method for generating a diffusion image of a nickel-based high-temperature alloy microstructure provided in the embodiment of the present application can be applied to Figure 1In the application environment shown. Among them, the terminal 102 communicates with the server 104 through the network. The data storage system can store the data that the server 104 needs to process. The data storage system can be set up separately, integrated on the server 104, or placed on the cloud or other servers. The terminal 102 can send the sample set and the original image of the nickel-based high-temperature alloy specimen to the server 104. After the server 104 receives the sample set and the original image of the nickel-based high-temperature alloy specimen, for the sample set and the original image of the nickel-based high-temperature alloy specimen, the server 104 uses the forward diffusion process of the denoising diffusion generation network to gradually diffuse the original training image into a Gaussian noise image; the reverse diffusion process of the denoising diffusion generation network is used to gradually denoise the Gaussian noise image to obtain a nickel-based high-temperature alloy microstructure image; determine the geometric structure characteristics of the nickel-based high-temperature alloy microstructure image The feature information is obtained, and the geometric feature difference is determined based on the geometric feature information corresponding to the original training image and the geometric feature information of the microstructure structure image of the nickel-based high-temperature alloy; the geometric feature loss penalty mechanism is used to determine whether the geometric feature difference is close to zero, and a trained denoising diffusion generation network is obtained; the reverse diffusion process of the trained denoising diffusion generation network is used to generate a virtual microstructure image based on the original image of the nickel-based high-temperature alloy specimen; the virtual microstructure image is screened using the grain boundary integrity principle to obtain a microstructure image of the nickel-based high-temperature alloy specimen. The server 104 can feed back the final microstructure image to the terminal 102. In addition, in some embodiments, the method for generating the diffusion image of the microstructure structure of the nickel-based high-temperature alloy may also be implemented independently by the server 104 or the terminal 102. For example, the terminal 102 may directly perform image diffusion and generation processing on the original images of the sample set and the nickel-based high-temperature alloy specimens, or the server 104 may obtain the original images of the sample set and the nickel-based high-temperature alloy specimens from the data storage system, and perform image diffusion and generation processing on the original images of the sample set and the nickel-based high-temperature alloy specimens.

[0060] The terminal 102 may be, but is not limited to, various desktop computers, laptop computers, smart phones, tablet computers, IoT devices, and portable wearable devices. The IoT devices may be smart speakers, smart TVs, smart air conditioners, smart vehicle-mounted devices, etc. The portable wearable devices may be smart watches, smart bracelets, head-mounted devices, etc. The server 104 may be implemented as an independent server or a server cluster consisting of multiple servers, or may be a cloud server.

[0061] In an exemplary embodiment, Figure 2As shown, a method for generating a diffusion image of a microstructure of a nickel-based high-temperature alloy is provided. The method is executed by a computer device, and can be executed by a computer device such as a terminal or a server alone, or by a terminal and a server together. In the embodiment of the present application, the method is applied to Figure 1 The server 104 in the example is used as an example to illustrate, including the following steps 200 to 209. Among them:

[0062] Step 200: Acquire multiple original samples of nickel-based high-temperature alloys to form a sample set. Each original sample of the nickel-based high-temperature alloy includes: an original training image and geometric structure feature information corresponding to the original training image. The original training image is a microstructure image of the nickel-based high-temperature alloy. The geometric structure feature information includes grain size, grain roundness, and grain major-minor axis ratio.

[0063] Step 201: The original training image is gradually diffused into a Gaussian noise-like image using the forward diffusion process of the denoising diffusion generation network.

[0064] Step 202: The Gaussian noise-like image is gradually denoised using the reverse diffusion process of the denoising diffusion generation network to obtain a microstructure image of the nickel-based high-temperature alloy.

[0065] Step 203: determine the geometric structure feature information of the nickel-based high-temperature alloy microstructure image, and determine the geometric feature difference based on the geometric structure feature information corresponding to the original training image and the geometric structure feature information of the nickel-based high-temperature alloy microstructure image.

[0066] Step 204: Determine whether the geometric feature difference is zero.

[0067] Step 205: When the geometric feature difference is not zero, the geometric feature difference is introduced into the denoising diffusion model as a geometric structure feature loss penalty term to obtain a trained denoising diffusion generation network.

[0068] Step 206: When the geometric feature difference is zero, zero is introduced into the denoising diffusion model as a geometric structure feature loss penalty term to obtain a trained denoising diffusion generation network.

[0069] Step 207: Acquire the original image of the nickel-based high-temperature alloy specimen. The original image of the nickel-based high-temperature alloy specimen is a microstructure image containing Gaussian noise.

[0070] Step 208: Generate a virtual microstructure image based on the original image of the nickel-based high-temperature alloy specimen by using the reverse diffusion process of the trained denoising diffusion generation network.

[0071] Step 209: Using the grain boundary integrity principle to screen the virtual microstructure image, and obtain the microstructure image of the nickel-based high-temperature alloy specimen.

[0072] In actual applications, multiple original samples of nickel-based high-temperature alloys and original images of nickel-based high-temperature alloy specimens can be obtained using an optical microscope.

[0073] The implementation of the above steps 200 to 209 can solve the problems faced by the prior art, such as the scarcity of data samples and the cumbersome, time-consuming and costly process of obtaining the internal microstructure of the alloy. It can intuitively reproduce similar microstructure images, thus bringing convenience to the study of material properties.

[0074] In another exemplary embodiment of the present application, in order to fully simulate the diffusion process of noise, the implementation process of the above step 201 can be: using the forward diffusion process of the denoising diffusion generation network, adding Gaussian noise to the original training image according to the set time step, to obtain a Gaussian noise-like image.

[0075] For example, the implementation process of step 201 given above is described by taking the Markov process as the forward diffusion process of the denoising diffusion generation network as an example, wherein:

[0076] Noise is added to the original training image in a Markov process. t |x t-1 ) is a conditional probability that defines the probability distribution of the image at step t where the noise is larger, given the image at step t-1 where the noise is smaller. That is, from the microstructure image x that follows the data distribution q(x) 0 Initially, the data x with less noise t-1 The calculation formula is as follows:

[0077]

[0078] in, represents the weight of the original training image to the noisy microstructure image after adding noise, N represents Gaussian distribution, and I represents the unit matrix. After reparameterization, the noisy microstructure image x t It can be written as:

[0079]

[0080] Among them, ε represents the added Gaussian noise.

[0081] After adding noise for T time steps, a Gaussian noise image is obtained.

[0082] In another exemplary embodiment of the present application, in order to gradually denoise the Gaussian noise image and restore the image similar to the original training microstructure image by using the reverse diffusion process of the denoising diffusion generation network, the implementation process of the above step 202 may be:

[0083] Step 2021: Input the Gaussian noise-like image and the set time step into the U-Net prediction network to obtain the added noise (i.e., the Gaussian noise to be added). For example, the U-Net network is used to predict the added noise for the Gaussian noise-like image.

[0084] Step 2022: Using the reverse diffusion process of the denoising diffusion generation network, remove the noise in the Gaussian noise-like image according to the set time step to obtain a microstructure image of the nickel-based high-temperature alloy. For example, using the Bayesian theorem, the denoising formula for restoring to a version with less noise is as follows:

[0085]

[0086] In the formula, μ θ (x t ,t) is the predicted mean, is the variance diffusion table, I represents the unit matrix, p θ (x t-1 |x t ) represents the reverse diffusion kernel.

[0087] Among them, the predicted mean μ θ (x t ,t) can be obtained by adding noise, and the calculation formula is as follows:

[0088]

[0089] In the formula, β t represents the variance, α t =1-β t Represented as a decreasing sequence, α t With β t The growth direction is opposite, ε θ (x t ,t) represents the added noise.

[0090] The Gaussian noise-like image is iteratively denoised by the predicted noise-added noise to restore an image similar to the original training microstructure image, i.e., a microstructure image of a nickel-based high-temperature alloy.

[0091] Based on the above description, the implementation principle architecture of the diffusion generation method of the nickel-based high-temperature alloy microstructure image provided in this application is as follows Figure 4 shown.

[0092] In another exemplary embodiment of the present application, in the process of using the geometric structure feature loss penalty mechanism to determine whether the geometric feature difference is zero, an image processing method can be used to obtain the geometric structure feature information of the original training image and the nickel-based high-temperature alloy microstructure image, and then the geometric feature difference is determined based on the obtained geometric structure feature information. Based on this, Figure 6 As shown, the process of determining geometric structure feature information can be described as:

[0093] (1) The grain area Area is determined by the contour area calculation method, and the grain size of a single grain is determined based on the grain area. The grain size is recorded as D. The calculation formula of the grain size D is:

[0094] (2) The grain length Len is determined by the contour length calculation method, and the roundness of a single grain is determined based on the grain area and grain length, which is recorded as M. The calculation formula for the roundness M is:

[0095] (3) The ellipse focus coordinate calculation method is used to determine the long axis (denoted as a) and short axis (denoted as b) of the grain in the microstructure image of the nickel-based high-temperature alloy. The long-short axis ratio of a single grain is determined based on the long axis and short axis of the grain, denoted as R. The calculation formula for the long-short axis ratio R is:

[0096] (4) Based on the grain size of a single grain, the roundness of a single grain, and the major-minor axis ratio of a single grain, the average grain size, the average roundness, and the average major-minor axis ratio are obtained.

[0097] (5) The average grain size, average roundness and average major-minor axis ratio are used as geometric structure feature information.

[0098] Furthermore, in this embodiment, the mean square error MSE calculation formula is used Calculate the geometric feature difference, that is, the mean square error is used as the geometric feature difference. i and i Represents the geometric structural feature information of the original training image and the microstructure image of the nickel-based high-temperature alloy.

[0099] In another exemplary embodiment of the present application, the virtual microscopic tissue structure image obtained in step 208 is obtained through training of a forward diffusion process and a reverse diffusion process, which includes:

[0100] Randomly generate a Gaussian noise image of the specified size.

[0101] The Gaussian noise image is input into the trained denoising diffusion generation network for the reverse diffusion process.

[0102] The required virtual microstructure image is output, and the virtual microstructure image is a two-dimensional image of the microstructure of the nickel-based high-temperature alloy.

[0103] In another exemplary embodiment of the present application, the virtual microstructure image can be screened using the integrity principle of the grain boundary in the real microstructure. The grain boundary integrity principle is that the grains in the microstructure must be closed intervals. The closed interval of the grain is identified by image processing break detection. Based on this, the implementation process of the above step 209 can be:

[0104] (1) Binarization is performed on the virtual microstructure image to obtain a binary image. For example, the virtual microstructure pixels are only 0 and 255, where 0 represents the black boundary line and 255 represents the inside of the grain.

[0105] (2) By means of image traversal, each pixel in the binary image is traversed, and the 8-neighborhood detection method is used to detect whether there is a 0 value near each 0-value pixel in the binary image. The vicinity of each 0-value pixel refers to the area extending from the 0-value pixel as the starting point to the set value.

[0106] (3) When there is no 0 value near each 0-value pixel, it indicates that there is a break in the grain boundary line, and the virtual microstructure image is deleted.

[0107] (4) When there is no 0 value near each 0-value pixel, it indicates that the grain boundary in the entire microstructure is intact, and the virtual microstructure image is retained, and the retained virtual microstructure image is used as the microstructure image of the nickel-based high-temperature alloy specimen.

[0108] In another exemplary embodiment of the present application, the microstructure image of GH4169 obtained by the diffusion generation method of the microstructure image of the nickel-based high-temperature alloy provided by the present application is taken as an example for explanation, wherein:

[0109] (1) Obtain the microstructure image of GH4169 and the corresponding geometric structure feature information through optical microscopy, and evaluate its effectiveness.

[0110] Optical microscope was used to collect microstructure images of GH4169 under different heat treatment processes.

[0111] The quality of the acquired GH4169 microstructure images was evaluated, and the microstructure images with poor performance were eliminated.

[0112] (2) Preprocessing the microstructure image obtained in step (1) to eliminate interference such as grain boundary blur, precipitated phase and polishing impurities.

[0113] (3) By using image grayscale conversion, image brightness contrast adjustment, grain boundary extraction and enhancement, etc., a pure and clear microstructure image is obtained and its geometric structure characteristics are calculated.

[0114] (4) Performing a weighted combination of the microstructure image obtained in step (3) and Gaussian noise to obtain a noisy microstructure image.

[0115] The noisy microstructure image continues to be weighted combined with Gaussian noise, the combination time step is T times, and the weight of Gaussian noise increases with the increase of time step.

[0116] (5) Noise prediction is performed on the noisy microstructure image that has undergone T times of noise addition.

[0117] The noisy microstructure image and the current time step T are used as the input of the U-Net prediction network to predict the noise.

[0118] (6) The predicted noise is removed from the noisy microstructure image to restore the original noise-free microstructure image.

[0119] (7) Using image processing methods, the geometric structure features inside the restored microstructure image are calculated, and the difference between the geometric structure features and the input original microstructure geometric features is calculated.

[0120] (8) A geometric structure feature loss penalty mechanism is used to determine whether the geometric feature difference is close to (or equal to) zero. When the geometric feature difference is not close to zero, the difference is introduced into the loss penalty term of the denoising diffusion model.

[0121] (9) A Gaussian noise image of a specified size is randomly generated and input into the reverse diffusion process of the trained denoising diffusion generation network to generate a virtual GH4169 microstructure image.

[0122] (10) The grain boundaries in the generated virtual GH4169 microstructure image are closed and the microstructure image with complete grain boundaries is selected as the final output.

[0123] As a specific example, Figure 3 As shown, the specific implementation process is as follows:

[0124] K1: Original sample collection and validity assessment, including:

[0125] K11: The microstructure images of GH4169 were obtained by optical microscope under different heat treatment processes. For example, the solution treatment was carried out at 960℃, the holding time was 60min, and then air cooling was performed, and the temperature gradient was 20℃. Then the temperature was kept at 720℃ for 8h, and then furnace cooled to 620℃ at a rate of 50℃ / h, and finally kept at the temperature for 8h and air cooled. Then the optical microscope was used to collect the images as shown below. Figure 5 Microstructure images are shown.

[0126] K12: Evaluate the quality of the acquired GH4169 microstructure images and remove the microstructure images with poor performance.

[0127] K2: Preprocessing of original microstructure images, including:

[0128] K21: Image grayscale processing is used to convert the original image into an 8-bit image, and brightness and contrast are adjusted to obtain an enhanced image.

[0129] K22: Select appropriate parameters to select and segment the region of interest (ROI) of the grains, and compare the original metallographic image to selectively ignore the edge grains of the image and the extremely small grains that were misselected, so as to improve the accuracy of grain identification and finally obtain a pure and clear microstructure image.

[0130] K23: Calculate the geometric structure characteristics of the segmented grains respectively to obtain the geometric structure characteristics of the overall image.

[0131] K3: The original microstructure image is processed by adding noise and converted into a noisy microstructure image, including:

[0132] K31: The noisy microstructure image continues to be weighted combined with Gaussian noise for T time steps, and the weight of Gaussian noise increases with the increase of time step.

[0133] K32: Noise prediction for a noisy microstructure image that has undergone T times of noise addition.

[0134] K4: U-Net network predicts the noise and restores the noisy microstructure image to a noise-free microstructure image, including:

[0135] K41: Use the noisy microstructure image and the current time step T as the input of the U-Net prediction network to predict the noise.

[0136] K42: The original noise-free microstructure image is restored by removing the predicted noise from the noisy microstructure image.

[0137] K5: Calculate the geometric features of the noise-free microstructure image and use the geometric features to determine the loss penalty, including:

[0138] K51: Use image processing, for details, refer to Figure 6 , calculate the geometric structure features inside the restored microstructure image, and calculate the difference between the geometric structure features and the input original microstructure geometric features (ie, the geometric feature difference).

[0139] K52: A geometric structure feature loss penalty mechanism is used to determine whether the geometric feature difference is zero. When the geometric feature difference is not zero, the difference is introduced into the denoising diffusion model as a geometric structure feature loss penalty term. When the geometric feature difference is zero, the geometric structure feature loss penalty term is zero.

[0140] K6: Use the trained denoising diffusion generation network (including geometric structure feature loss penalty) to obtain virtual microstructure images, including:

[0141] K61: Randomly generates a Gaussian noise image of a specified size.

[0142] K62: The Gaussian noise image is input into the reverse diffusion process of the trained denoising diffusion generation network to generate a virtual two-dimensional image of the microstructure of the nickel-based high-temperature alloy.

[0143] K7: Perform grain boundary closure detection on the generated microstructure image, including:

[0144] K71: Binarize the generated microstructure image.

[0145] K72: Use image traversal to traverse every pixel in the microstructure image.

[0146] K73: The 8-neighborhood detection method is used to detect whether there is one or more 0 values ​​near each 0-value pixel. If not, it indicates that the boundary line is broken, otherwise the grain boundary in the entire microstructure is intact.

[0147] Based on the same inventive concept, the embodiment of the present application also provides a nickel-based high-temperature alloy microstructure image diffusion generation system for realizing the nickel-based high-temperature alloy microstructure image diffusion generation method involved above. The implementation scheme for solving the problem provided by the system is similar to the implementation scheme recorded in the above method, so the specific limitations in one or more nickel-based high-temperature alloy microstructure image diffusion generation system embodiments provided below can refer to the limitations of the nickel-based high-temperature alloy microstructure image diffusion generation method above, and will not be repeated here.

[0148] In an exemplary embodiment, Figure 7 As shown, a nickel-based high-temperature alloy microstructure image diffusion generation system is provided, comprising:

[0149] The data acquisition module is used to acquire multiple original samples of nickel-based superalloys to form a sample set, and to acquire the original image of the nickel-based superalloy specimen. Each original sample of the nickel-based superalloy includes: an original training image and geometric structure feature information corresponding to the original training image. The original training image is a microstructure image of the nickel-based superalloy. The geometric structure feature information includes grain size, grain roundness, and the major-minor axis ratio of the grain. The original image of the nickel-based superalloy specimen is a microstructure image containing Gaussian noise.

[0150] The forward diffusion module is used to gradually diffuse the original training image into a Gaussian noise-like image using the forward diffusion process of the denoising diffusion generation network.

[0151] The reverse diffusion module is used to gradually denoise the Gaussian noise image by using the reverse diffusion process of the denoising diffusion generation network to obtain the microstructure image of the nickel-based high-temperature alloy.

[0152] The geometric feature difference determination module is used to determine the geometric structure feature information of the nickel-based high-temperature alloy microstructure image, and determine the geometric feature difference based on the geometric structure feature information corresponding to the original training image and the geometric structure feature information of the nickel-based high-temperature alloy microstructure image.

[0153] The geometric structure feature loss penalty module is used to determine whether the geometric feature difference is zero. When the geometric feature difference is not zero, the geometric feature difference is introduced into the denoising diffusion model as a geometric structure feature loss penalty term to obtain a trained denoising diffusion generation network. When the geometric feature difference is zero, zero is introduced into the denoising diffusion model as a geometric structure feature loss penalty term to obtain a trained denoising diffusion generation network.

[0154] The microstructure generation module is used to generate a virtual microstructure image based on the original image of the nickel-based high-temperature alloy specimen by using the reverse diffusion process of the trained denoising diffusion generation network.

[0155] The microstructure image generation module is used to screen the virtual microstructure image by adopting the grain boundary integrity principle to obtain the microstructure image of the nickel-based high-temperature alloy specimen.

[0156] As an optional implementation, in practical applications, such as Figure 8 As shown, the nickel-based high-temperature alloy microstructure image diffusion generation system may also include:

[0157] The data collection and evaluation module is used to collect GH4169 microstructure images and corresponding geometric structure feature information, and to evaluate the effectiveness of all acquired data and eliminate microstructure images with poor performance.

[0158] The microstructure image preprocessing module is used to eliminate interference such as grain boundary blur, precipitated phase and polishing impurities to obtain pure and clear microstructure images.

[0159] The microstructure image noise adding module destroys the structural features of the original image by adding Gaussian noise to obtain a noisy microstructure image.

[0160] The noise prediction and denoising module predicts the added noise through the U-Net network, and uses the added noise to remove the noise inside the noisy microstructure image to obtain a noise-free microstructure image.

[0161] The geometric feature penalty module is used to determine whether there is a difference between the geometric structure features of the restored noise-free microstructure image and the geometric structure features of the original training microstructure. If there is a difference, it indicates that the generation effect is not good and the model needs to be penalized.

[0162] The microscopic tissue structure generation module is used to generate a virtual two-dimensional image of the microscopic tissue structure by using the module alone after the training is completed.

[0163] The closure detection module is used to determine whether the internal grains of the generated two-dimensional image of the microstructure conform to objective laws and whether the grain boundaries of the internal grains are closed small areas.

[0164] In this embodiment, the data collection and evaluation module can collect microstructure images of nickel-based high-temperature alloys obtained by optical microscope under different heat treatment processes, evaluate the quality of the obtained microstructure images, and eliminate microstructure images with poor performance.

[0165] The microstructure image preprocessing module can eliminate interferences such as grain boundary blur, precipitated phases and polishing impurities. Image grayscale processing is used to adjust brightness and contrast to obtain an enhanced image. The original metallographic image is compared to selectively ignore the edge grains of the image and the extremely small grains that are misselected, so as to improve the accuracy of grain recognition and finally obtain a pure and clear microstructure image. The geometric structure characteristics of the identified grains are calculated separately to obtain the geometric structure characteristics of the overall image.

[0166] The microstructure image noise module can generate random Gaussian noise, and transform the original microstructure image into a noisy microstructure image through linear weighted processing. The noisy microstructure image is further weighted combined with Gaussian noise for T times, and the weight of Gaussian noise increases with the increase of time step.

[0167] The noise prediction and denoising module can be used for the U-Net network of noise prediction to predict the added noise in the denoising process, obtain the distribution of the denoised microstructure image with the help of the Bayesian formula, and restore the original noise-free microstructure image through sampling.

[0168] The geometric feature penalty module can use the microstructure geometric feature calculation formula to obtain the geometric structure features of the restored microstructure image, calculate the mean square error between the geometric structure features and the geometric structure features corresponding to the input original microstructure image, and use the mean square error as one of the loss penalties of the prediction network U-Net to optimize the weight of the prediction network.

[0169] The microstructure generation module can randomly generate a Gaussian noise image of the same size as the microstructure image to be generated, and use the Gaussian noise image as the input of the denoising process, and gradually denoise it to obtain a virtual microstructure image.

[0170] The closure detection module can perform binarization processing on the virtual microstructure image generated by the microstructure generation module, detect whether there are unclosed broken lines at the grain boundaries inside, and screen out microstructure images containing unclosed broken lines.

[0171] In summary, the present application obtains the optical microstructure image of the nickel-based high-temperature alloy and evaluates the validity of all the original sample data obtained; then pre-processes the above microstructure image to eliminate interference such as grain boundary blur, precipitated phase and polishing impurities, and calculates the corresponding geometric structure features; uses the forward diffusion process to add noise to the original microstructure; denoises the noisy microstructure image through the reverse diffusion process, and uses the mean square error between the geometric structure features of the denoised microstructure image and the geometric structure features of the input original microstructure as the loss penalty term of the model, which is used to optimize the prediction model; reconstructs the virtual microstructure through the trained model; performs closure detection on the obtained virtual microstructure image to screen out microstructure images that do not meet the requirements. This application aims to solve the problem of the difficulty in obtaining existing microstructures, and by introducing a deep learning model, it can achieve convenient and fast microstructure image acquisition.

[0172] In an exemplary embodiment, a computer device is provided. The computer device may be a server or a terminal. The internal structure diagram thereof may be as follows: Fig. 9 As shown. The computer device includes a processor, a memory, an input / output interface (Input / Output, referred to as I / O) and a communication interface. Among them, the processor, the memory and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store image diffusion and generate processing data. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a method for generating a diffusion image of a microstructure structure of a nickel-based high-temperature alloy is implemented.

[0173] Those skilled in the art will understand that Fig. 9 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components. In an exemplary embodiment, a computer device is provided, including a memory and a processor, wherein a computer program is stored in the memory, and the processor implements the steps in the above-mentioned method embodiments when executing the computer program.

[0174] In an exemplary embodiment, a computer-readable storage medium is provided, storing a computer program, and when the computer program is executed by a processor, the steps in the above method embodiments are implemented.

[0175] In an exemplary embodiment, a computer program product is provided, including a computer program, and when the computer program is executed by a processor, the steps in the above method embodiments are implemented.

[0176] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant regulations.

[0177] Those of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in the present 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 may be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM).

[0178] The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. The non-relational database may include a distributed database based on blockchain, etc., but is not limited thereto. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., but is not limited thereto.

[0179] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0180] This article uses specific examples to illustrate the principles and implementation methods of this application. The description of the above embodiments is only used to help understand the method and core ideas of this application. At the same time, for those skilled in the art, according to the ideas of this application, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting this application.

Claims

1. A method for generating a diffusion image of a nickel-based high-temperature alloy microstructure, characterized in that: The method for generating a diffusion-generated image of a microstructure of a nickel-based high-temperature alloy comprises: Acquire multiple original samples of nickel-based high-temperature alloys to form a sample set; each original sample of the nickel-based high-temperature alloy includes: an original training image and geometric structure feature information corresponding to the original training image; the original training image is a microstructure image of the nickel-based high-temperature alloy; the geometric structure feature information includes grain size, grain roundness, and major-minor axis ratio of the grain; The original training image is gradually diffused into a Gaussian noise-like image by using a forward diffusion process of a denoising diffusion generation network; The Gaussian noise-like image is gradually denoised by using the reverse diffusion process of the denoising diffusion generation network to obtain a microstructure image of the nickel-based high-temperature alloy; Determine the geometric structure feature information of the nickel-based high-temperature alloy microstructure image, and determine the geometric feature difference based on the geometric structure feature information corresponding to the original training image and the geometric structure feature information of the nickel-based high-temperature alloy microstructure image; Determining whether the geometric feature difference is zero; When the geometric feature difference is not zero, the geometric feature difference is introduced into the denoising diffusion model as a geometric structure feature loss penalty term to obtain a trained denoising diffusion generation network; When the geometric feature difference is zero, zero is introduced into the denoising diffusion model as a geometric structure feature loss penalty term to obtain a trained denoising diffusion generation network; Acquire an original image of a nickel-based high-temperature alloy specimen; the original image of the nickel-based high-temperature alloy specimen is a microstructure image containing Gaussian noise; Generate a virtual microstructure image based on the original image of the nickel-based high-temperature alloy specimen by using a reverse diffusion process of a trained denoising diffusion generation network; The virtual microstructure image is screened using the grain boundary integrity principle to obtain the microstructure image of the nickel-based high-temperature alloy specimen.

2. The method for generating a diffusion image of a nickel-based high-temperature alloy microstructure according to claim 1, characterized in that: The forward diffusion process of the denoising diffusion generation network is used to gradually diffuse the original training image into a Gaussian noise-like image, specifically including: The forward diffusion process of the denoising diffusion generation network is adopted to add Gaussian noise to the original training image according to a set time step to obtain the Gaussian noise-like image.

3. The method for generating a diffusion image of a nickel-based high-temperature alloy microstructure according to claim 2, characterized in that: The forward diffusion process of the denoising diffusion generation network is a Markov process.

4. The method for generating a diffusion image of a nickel-based high-temperature alloy microstructure according to claim 2, characterized in that: The Gaussian noise image is gradually denoised by the reverse diffusion process of the denoising diffusion generation network to obtain a microstructure image of the nickel-based high-temperature alloy, specifically including: Inputting the Gaussian noise-like image and the set time step into a U-Net prediction network to obtain a noise-added noise; The reverse diffusion process of the denoising diffusion generation network is adopted to remove the noise in the Gaussian noise-like image according to the set time step to obtain the microstructure image of the nickel-based high-temperature alloy.

5. The method for generating a diffusion image of a nickel-based high-temperature alloy microstructure according to claim 1, characterized in that: Determining the geometric structure feature information of the nickel-based high-temperature alloy microstructure image specifically includes: Determine the grain area in the nickel-based high-temperature alloy microstructure image by using a contour area calculation method, and determine the grain size of a single grain based on the grain area; Determine the grain length in the nickel-based high-temperature alloy microstructure image by using a contour length calculation method, and determine the roundness of a single grain based on the grain area and the grain length; The major axis and the minor axis of the grains in the microstructure image of the nickel-based high-temperature alloy are determined by using an ellipse focus coordinate calculation method, and the major-minor axis ratio of a single grain is determined based on the major axis and the minor axis of the grain; Obtaining an average grain size, an average roundness, and an average major-minor axis ratio based on the grain size of a single grain, the roundness of a single grain, and the major-minor axis ratio of a single grain; The average grain size, average roundness and average major-minor axis ratio are used as geometric structural feature information of the microstructure image of the nickel-based high-temperature alloy.

6. The method for generating a diffusion image of a nickel-based high-temperature alloy microstructure according to claim 1, characterized in that: The virtual microstructure image is screened by adopting the grain boundary integrity principle to obtain the microstructure image of the nickel-based high-temperature alloy specimen, specifically including: Binarizing the virtual microscopic tissue structure image to obtain a binary image; By means of image traversal, each pixel in the binary image is traversed, and an 8-neighborhood detection method is used to detect whether there is a 0 value near each 0-value pixel in the binary image; the vicinity of each 0-value pixel refers to an area extending from the 0-value pixel as a starting point to a set value; When there is no 0 value near each 0-value pixel, it indicates that there is a break in the grain boundary line, and the virtual microstructure image is deleted; When there is no 0 value near each 0-value pixel, it indicates that the grain boundaries in the entire microstructure are intact, and the virtual microstructure image is retained, and the retained virtual microstructure image is used as the microstructure image of the nickel-based high-temperature alloy specimen.

7. A nickel-based high-temperature alloy microstructure image diffusion generation system, characterized in that: The nickel-based high-temperature alloy microstructure image diffusion generation system comprises: A data acquisition module, used to acquire multiple original samples of nickel-based high-temperature alloys to form a sample set, and to acquire original images of nickel-based high-temperature alloy specimens; each original sample of the nickel-based high-temperature alloy includes: an original training image and geometric structure feature information corresponding to the original training image; the original training image is a microstructure image of the nickel-based high-temperature alloy; the geometric structure feature information includes grain size, grain roundness, and major-minor axis ratio of the grain; the original image of the nickel-based high-temperature alloy specimen is a microstructure image containing Gaussian noise; A forward diffusion module, used for gradually diffusing the original training image into a Gaussian noise-like image by using a forward diffusion process of a denoising diffusion generation network; A reverse diffusion module is used to gradually denoise the Gaussian noise image by using a reverse diffusion process of a denoising diffusion generation network to obtain a microstructure image of a nickel-based high-temperature alloy; A geometric feature difference determination module, used to determine the geometric structure feature information of the nickel-based high-temperature alloy microstructure image, and determine the geometric feature difference based on the geometric structure feature information corresponding to the original training image and the geometric structure feature information of the nickel-based high-temperature alloy microstructure image; A geometric structure feature loss penalty module, used to determine whether the geometric feature difference is zero; when the geometric feature difference is not zero, the geometric feature difference is introduced into the denoising diffusion model as a geometric structure feature loss penalty term to obtain a trained denoising diffusion generation network; when the geometric feature difference is zero, zero is introduced into the denoising diffusion model as a geometric structure feature loss penalty term to obtain a trained denoising diffusion generation network; A microstructure generation module, for generating a virtual microstructure image based on an original image of the nickel-based high-temperature alloy specimen by using a reverse diffusion process of a trained denoising diffusion generation network; The microstructure image generation module is used to screen the virtual microstructure image by adopting the grain boundary integrity principle to obtain the microstructure image of the nickel-based high-temperature alloy specimen.

8. A computer device comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the method for generating a diffusion image of a microstructure of a nickel-based high-temperature alloy as described in any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method for generating a diffusion-generated image of a microstructure of a nickel-based high-temperature alloy according to any one of claims 1 to 6 is implemented.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the method for generating a diffusion-generated image of a microstructure of a nickel-based high-temperature alloy according to any one of claims 1 to 6 is implemented.

Images