Alloy microstructure identification method, device and computer equipment

Abstract

The present application relates to the construction of an alloy microstructure recognition model and a recognition method, device, computer equipment and storage medium. The method comprises: obtaining alloy microstructure images of different stresses and heat treatment states, and inputting the obtained γ' phase microstructure image set into a preset convolutional neural network model for training through image segmentation to obtain a convolutional neural network model of the alloy microstructure image. When performing application recognition, the obtained γ' phase microstructure image set to be tested is input into the convolutional neural network model of the alloy microstructure image to obtain the stress and heat treatment state linear classification results corresponding to the alloy microstructure image to be tested. The use of this method can improve the efficiency of alloy microstructure recognition.

Description

Technical Field

[0001] The present application relates to the field of image recognition technology, and in particular to an alloy microstructure recognition method, device, computer equipment, storage medium and computer program product. Background Art

[0002] Nickel-based single crystal high-temperature alloy is currently the highest-grade material for the preparation of aircraft engine blades. China and major developed countries attach great importance to the research and development of single crystal high-temperature alloys and the development and safe service of single crystal blades. Compared with cast equiaxed crystals and directionally solidified high-temperature alloys, the microstructure of single crystal high-temperature alloys is simpler. It is mainly composed of γ and precipitation strengthening phase γ' phase. The content and morphology of γ' phase have the most significant influence on the various properties of single crystal high-temperature alloys. Quantitative characterization of the content and morphology of γ' phase in single crystal high-temperature alloy blades is very important for inspecting the factory performance of blades, evaluating the safety and reliability of blade service, and predicting the remaining service life of blades.

[0003] At present, the traditional manual method is still used to characterize the microstructure of nickel-based single crystal high-temperature alloys. Software such as Photoshop, Image, and MATLAB are used to perform statistics on alloy microstructure photos, which has the problem of low image recognition efficiency. Summary of the invention

[0004] Based on this, it is necessary to provide a method, device, computer equipment, computer-readable storage medium and computer program product for constructing an alloy microstructure identification model that can improve image recognition efficiency, and a method, device, computer equipment, computer-readable storage medium and computer program product for identifying an alloy microstructure that can improve image recognition efficiency, in order to address the technical problem of low image recognition efficiency in traditional alloy microstructure quantitative statistical methods.

[0005] In a first aspect, the present application provides a method for constructing an alloy microstructure identification model. The method comprises:

[0006] Obtain microstructure images of alloys in different stress and heat treatment states;

[0007] According to the alloy microstructure image, a γ' phase microstructure image set is obtained by image segmentation;

[0008] The γ' phase microstructure image set is input into a preset convolutional neural network model for training to obtain a convolutional neural network model of the alloy microstructure image, wherein the training process includes extracting feature data of the γ' phase microstructure image set, classifying the feature data of the γ' phase microstructure image set through a linear activation function, and outputting the linear classification results of stress and heat treatment state corresponding to the γ' phase microstructure image set.

[0009] In one embodiment, obtaining a γ' phase microstructure image set by image segmentation based on the alloy microstructure image includes:

[0010] According to the alloy microstructure image, the initial γ' phase microstructure image set is obtained by image segmentation;

[0011] The initial γ' phase microstructure image set is subjected to a preprocessing method to obtain a γ' phase microstructure image set, and the preprocessing method includes a denoising algorithm and image normalization.

[0012] In one embodiment, the γ' phase microstructure image set is input into a preset convolutional neural network model for training, and the convolutional neural network model for obtaining the alloy microstructure image includes:

[0013] Inputting the γ' phase microstructure image set into the input layer of the preset convolutional neural network model, and obtaining feature data of the γ' phase microstructure image set through convolution and pooling operations in the hidden layer of the preset convolutional neural network model;

[0014] According to the characteristic data, the stress and heat treatment state test results corresponding to the characteristic data are obtained through the forward propagation function, and the test results are sent to the output layer of the preset convolutional neural network model;

[0015] According to the stress and heat treatment state test results, the parameters of the forward propagation function are updated through back propagation in the output layer to obtain the convolutional neural network model of the alloy microstructure image.

[0016] In one embodiment, the γ' phase microstructure image set is input into the input layer of the preset convolutional neural network model, and the feature data of the γ' phase microstructure image set is obtained by convolution and pooling operations in the hidden layer of the preset convolutional neural network model, including:

[0017] Inputting the γ' phase microstructure image set into the input layer of the preset convolutional neural network model, obtaining an activated image set through an activation function, and passing the activated image set to the hidden layer of the preset convolutional neural network model;

[0018] According to the activated image set, the feature vector is obtained through the convolution kernel and activation function in the hidden layer;

[0019] According to the feature vector, through the activation function and pooling operation, the pooled feature vector is obtained, and the pooling operation count value is accumulated;

[0020] After the pooled feature vector is activated by an activation function, the activated image set is updated, and the step of obtaining the feature vector by passing the convolution kernel and the activation function in the hidden layer according to the activated image set is returned;

[0021] If the accumulated count value of the pooling operation is greater than the pre-designed value, the feature data of the γ' phase microstructure image set is obtained according to the feature vector after pooling.

[0022] In one embodiment, according to the stress and heat treatment state test results, the convolutional neural network model for obtaining the alloy microstructure image by back propagation in the output layer and updating the parameters of the forward propagation function includes:

[0023] According to the γ' phase microstructure image set, the initial stress and heat treatment state data corresponding to the γ' phase microstructure image set are obtained;

[0024] A loss function is obtained according to the stress and heat treatment state test results and the initial stress and heat treatment state data;

[0025] If the loss function is greater than the preset threshold, the output layer updates the parameters of the forward propagation function by backpropagating and calculating the gradient;

[0026] If the loss function is less than or equal to the preset threshold, the feature data of the γ' phase microstructure image set is classified through a linear activation function in the output layer, the linear classification results of stress and heat treatment state corresponding to the alloy microstructure image are output, and a convolutional neural network model of the alloy microstructure image is constructed.

[0027] In a second aspect, the present application also provides a device for constructing an alloy microstructure identification model. The device comprises:

[0028] Image data acquisition module, used to obtain alloy microstructure images under different stress and heat treatment states;

[0029] An image segmentation module is used to obtain a γ' phase microstructure image set by image segmentation according to the alloy microstructure image;

[0030] The model training module is used to input the γ' phase microstructure image set into a preset convolutional neural network model for training to obtain a convolutional neural network model of the alloy microstructure image, wherein the training process includes extracting feature data of the γ' phase microstructure image set, classifying the feature data of the γ' phase microstructure image set through a linear activation function, and outputting the linear classification results of stress and heat treatment state corresponding to the γ' phase microstructure image set.

[0031] In a third aspect, the present application further provides a computer device. The computer device includes a memory and a processor, the memory stores a computer program, and the processor implements the following steps when executing the computer program:

[0032] Acquire alloy microstructure images of different stresses and heat treatment states; obtain a γ' phase microstructure image set through image segmentation based on the alloy microstructure image; input the γ' phase microstructure image set into a preset convolutional neural network model for training to obtain a convolutional neural network model of the alloy microstructure image, wherein the training process includes extracting feature data of the γ' phase microstructure image set, classifying the feature data of the γ' phase microstructure image set through a linear activation function, and outputting a linear classification result of the stress and heat treatment state corresponding to the γ' phase microstructure image set.

[0033] In a fourth aspect, the present application further provides a computer-readable storage medium. The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the following steps are implemented:

[0034] Acquire alloy microstructure images of different stresses and heat treatment states; obtain a γ' phase microstructure image set through image segmentation based on the alloy microstructure image; input the γ' phase microstructure image set into a preset convolutional neural network model for training to obtain a convolutional neural network model of the alloy microstructure image, wherein the training process includes extracting feature data of the γ' phase microstructure image set, classifying the feature data of the γ' phase microstructure image set through a linear activation function, and outputting a linear classification result of the stress and heat treatment state corresponding to the γ' phase microstructure image set.

[0035] In a fifth aspect, the present application further provides a computer program product. The computer program product includes a computer program, and when the computer program is executed by a processor, the following steps are implemented:

[0036] Acquire alloy microstructure images of different stresses and heat treatment states; obtain a γ' phase microstructure image set through image segmentation based on the alloy microstructure image; input the γ' phase microstructure image set into a preset convolutional neural network model for training to obtain a convolutional neural network model of the alloy microstructure image, wherein the training process includes extracting feature data of the γ' phase microstructure image set, classifying the feature data of the γ' phase microstructure image set through a linear activation function, and outputting a linear classification result of the stress and heat treatment state corresponding to the γ' phase microstructure image set.

[0037] The above-mentioned alloy microstructure recognition model construction method, device, computer equipment, storage medium and computer program product can improve the alloy microstructure recognition efficiency by inputting alloy microstructure images of different stresses and heat treatment states into a preset convolutional neural network model for training after image segmentation. Since the feature data of the γ' phase microstructure image set corresponds one-to-one to the stress and heat treatment state of the alloy microstructure image, the feature data of the γ' phase microstructure image set is classified through a linear activation function during the training process, and the linear classification results of the stress and heat treatment state corresponding to the γ' phase microstructure image set can be obtained.

[0038] In a sixth aspect, the present application provides a method for identifying alloy microstructures. The method comprises:

[0039] Obtaining microstructure images of the alloy to be tested;

[0040] According to the microstructure image of the alloy to be tested, a set of γ' phase microstructure images to be tested is obtained by image segmentation;

[0041] The γ' phase microstructure image set to be tested is input into the convolutional neural network model of the alloy microstructure image to obtain the linear classification results of stress and heat treatment state corresponding to the alloy microstructure image to be tested; wherein, the convolutional neural network model of the alloy microstructure image is trained using the alloy microstructure recognition model construction method.

[0042] In a seventh aspect, the present application also provides an alloy microstructure identification device. The device comprises:

[0043] The image data acquisition module to be tested is used to obtain the microstructure image of the alloy to be tested;

[0044] The image segmentation module to be tested is used to obtain a set of γ' phase microstructure images to be tested by image segmentation according to the microstructure image of the alloy to be tested;

[0045] The classification result acquisition module is used to input the γ' phase microstructure image set to be tested into the convolutional neural network model of the alloy microstructure image to obtain the linear classification results of stress and heat treatment state corresponding to the alloy microstructure image to be tested; wherein, the convolutional neural network model of the alloy microstructure image is trained using the alloy microstructure recognition model construction method.

[0046] In an eighth aspect, the present application further provides a computer device. The computer device includes a memory and a processor, the memory stores a computer program, and the processor implements the following steps when executing the computer program:

[0047] A microstructure image of the alloy to be tested is obtained; according to the microstructure image of the alloy to be tested, a set of γ' phase microstructure images to be tested is obtained by image segmentation; the set of γ' phase microstructure images to be tested is input into a convolutional neural network model of the alloy microstructure image to obtain linear classification results of stress and heat treatment state corresponding to the microstructure image of the alloy to be tested; wherein the convolutional neural network model of the alloy microstructure image is trained using an alloy microstructure recognition model construction method.

[0048] In a ninth aspect, the present application further provides a computer-readable storage medium. The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the following steps are implemented:

[0049] A microstructure image of the alloy to be tested is obtained; according to the microstructure image of the alloy to be tested, a set of γ' phase microstructure images to be tested is obtained by image segmentation; the set of γ' phase microstructure images to be tested is input into a convolutional neural network model of the alloy microstructure image to obtain linear classification results of stress and heat treatment state corresponding to the microstructure image of the alloy to be tested; wherein the convolutional neural network model of the alloy microstructure image is trained using an alloy microstructure recognition model construction method.

[0050] In a tenth aspect, the present application further provides a computer program product. The computer program product includes a computer program, and when the computer program is executed by a processor, the following steps are implemented:

[0051] A microstructure image of the alloy to be tested is obtained; according to the microstructure image of the alloy to be tested, a set of γ' phase microstructure images to be tested is obtained by image segmentation; the set of γ' phase microstructure images to be tested is input into a convolutional neural network model of the alloy microstructure image to obtain linear classification results of stress and heat treatment state corresponding to the microstructure image of the alloy to be tested; wherein the convolutional neural network model of the alloy microstructure image is trained using an alloy microstructure recognition model construction method.

[0052] The above-mentioned alloy microstructure identification method, device, computer equipment, storage medium and computer program product obtain the linear classification results of stress and heat treatment state corresponding to the alloy microstructure image to be tested by inputting the image into the convolutional neural network model of the alloy microstructure image after image segmentation, wherein the convolutional neural network model of the alloy microstructure image is trained using the alloy microstructure identification model construction method. This method of alloy microstructure identification using the convolutional neural network model improves the efficiency of alloy microstructure identification. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] Figure 1 is an application environment diagram of an alloy microstructure identification method in an embodiment;

[0054] Figure 2 A schematic diagram of a process for constructing an alloy microstructure identification model in one embodiment;

[0055] Figure 3 A schematic diagram of a process for constructing an alloy microstructure identification model in one embodiment;

[0056] Figure 4 Schematic diagram of a sub-process of S420 in one embodiment;

[0057] Figure 5 is a schematic diagram of a sub-process of S460 in one embodiment;

[0058] Figure 6 A structural block diagram of an alloy microstructure identification model building device in one embodiment;

[0059] Figure 7 A schematic diagram of a process of an alloy microstructure identification method in one embodiment;

[0060] Figure 8 is a structural block diagram of an alloy microstructure identification device in one embodiment;

[0061] Fig. 9 FIG. 4 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0062] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0063] The alloy microstructure identification method provided in the embodiment of the present application can be applied to Figure 1In the application environment shown, the terminal 102 communicates with the server 104 through a network. The data storage system can store data that the server 104 needs to process. The data storage system can be integrated on the server 104, or placed on the cloud or other network servers. Terminal 102 obtains alloy microstructure images of different stresses and heat treatment states; obtains a γ' phase microstructure image set by image segmentation based on the alloy microstructure image; inputs the γ' phase microstructure image set into a preset convolutional neural network model for training to obtain a convolutional neural network model of the alloy microstructure image, wherein the training process includes extracting feature data of the γ' phase microstructure image set, classifying the feature data of the γ' phase microstructure image set by a linear activation function, and outputting linear classification results of stress and heat treatment state corresponding to the γ' phase microstructure image set; obtains the alloy microstructure image to be tested; obtains a γ' phase microstructure image set to be tested by image segmentation based on the alloy microstructure image to be tested; inputs the γ' phase microstructure image set to be tested into the convolutional neural network model of the alloy microstructure image, and obtains linear classification results of stress and heat treatment state corresponding to the alloy microstructure image to be tested; wherein the convolutional neural network model of the alloy microstructure image is trained using the alloy microstructure recognition model construction method. The terminal 102 may be, but is not limited to, various personal computers, laptops, smart phones, tablet computers, IoT devices, and portable wearable devices. The IoT devices may be smart speakers, smart TVs, smart air conditioners, smart car-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.

[0064] In one embodiment, Figure 2 As shown in the figure, a method for constructing an alloy microstructure identification model is provided, and the method is applied to Figure 1 The terminal 102 in the example is used as an example to illustrate, and the following steps are included:

[0065] S200, obtain alloy microstructure images under different stress and heat treatment conditions.

[0066] Among them, when the alloy is deformed due to external factors, the internal forces that interact with each other are generated between the parts of the object to resist the action of this external factor and try to restore the alloy from the position after deformation to the position before deformation. This internal force per unit area at a certain point in the cross section under investigation is called stress. Heat treatment refers to a metal heat processing process in which the material is heated, kept warm and cooled in the solid state to obtain the expected structure and performance. Heat treatment at different temperatures is the state of heat treatment. Under different stress and heat treatment states, the alloy microstructure image is different.

[0067] Specifically, microstructure images of alloys under different stress and heat treatment states were obtained.

[0068] S300, obtaining a γ' phase microstructure image set by image segmentation according to the alloy microstructure image.

[0069] Among them, the microstructure of single crystal high-temperature alloy is mainly composed of γ phase and γ' phase. The content and morphology of γ' phase have the most significant influence on the various properties of single crystal high-temperature alloy. Image segmentation is to select different areas on the alloy test sample, collect a photo for each different area, and obtain different image sets. Each image in the image set includes γ' phase particles. The image set containing γ' phase particles is the γ' phase microstructure image set.

[0070] Specifically, according to the alloy microstructure image, a set of γ' phase microstructure images is obtained through image segmentation.

[0071] S400, inputting the γ' phase microstructure image set into a preset convolutional neural network model for training, and obtaining a convolutional neural network model of the alloy microstructure image, wherein the training process includes extracting feature data of the γ' phase microstructure image set, classifying the feature data of the γ' phase microstructure image set through a linear activation function, and outputting a linear classification result of stress and heat treatment state corresponding to the γ' phase microstructure image set.

[0072] Among them, the convolutional neural network is a type of feedforward neural network that includes convolution calculations and has a deep structure. It is one of the representative algorithms of deep learning. The structure of the convolutional neural network includes an input layer, a hidden layer, and an output layer. The classic convolutional neural network model types include: LeNet, AlexNet, VGG, NiN, GooLeNet, ResNet, and DenseNet. This application does not specifically limit the type of convolutional neural network. The activation function is a function that runs on the neurons of the artificial neural network, responsible for mapping the input of the neuron to the output end. The linear activation function is the simplest activation function. The output is proportional to the input, its derivative is a constant, and the gradient is also a constant. The γ' phase in the single crystal high-temperature alloy blade before service maintains an approximately cubic morphology. During the service of the blade, under the action of temperature and stress, the γ' phase will undergo various content and morphological changes such as dissolution, coarsening connection, and rafting. Therefore, the content and morphology of the γ' phase can reflect the stress and heat treatment state of the alloy.

[0073] Specifically, the γ' phase microstructure image set is input into a preset convolutional neural network model for training to obtain a convolutional neural network model of the alloy microstructure image, wherein the training process includes extracting feature data of the γ' phase microstructure image set, classifying the feature data of the γ' phase microstructure image set through a linear activation function, and outputting the linear classification results of stress and heat treatment state corresponding to the γ' phase microstructure image set.

[0074] In the above-mentioned alloy microstructure recognition model construction method, the alloy microstructure images of different stresses and heat treatment states are input into a preset convolutional neural network model for training after image segmentation to obtain a convolutional neural network model of the alloy microstructure image, which can improve the alloy microstructure recognition efficiency. Since the feature data of the γ' phase microstructure image set corresponds one-to-one to the stress and heat treatment state of the alloy microstructure image, the feature data of the γ' phase microstructure image set is classified through a linear activation function during the training process, and the linear classification results of the stress and heat treatment state corresponding to the γ' phase microstructure image set can be obtained.

[0075] In one embodiment, obtaining a γ' phase microstructure image set through image segmentation based on an alloy microstructure image includes: obtaining an initial γ' phase microstructure image set through image segmentation based on an alloy microstructure image; obtaining a γ' phase microstructure image set through a preprocessing method for the initial γ' phase microstructure image set, the preprocessing method including a denoising algorithm and image normalization.

[0076] In this embodiment, different areas are selected on the alloy test sample, and a photo is collected for each different area to obtain different image sets, each image in the image set includes γ' phase particles, and the image set containing γ' phase particles is obtained by image segmentation according to the alloy microstructure images under different stresses and heat treatment states as the initial γ' phase microstructure image set. The initial γ' phase microstructure image set is subjected to a denoising algorithm to obtain a denoised image set, wherein the denoising algorithm used in this application includes but is not limited to color gamut conversion, median filtering, edge detection, grayscale and binarization, and this application does not limit the type of denoising algorithm, and the denoised image set is subjected to an image normalization operation, and the image normalization operation can convert the image set into an image set with the same pixel having the same grayscale range, and the image set obtained after normalization is the γ' phase microstructure image set.

[0077] The scheme of the above-mentioned embodiment obtains an initial γ' phase microstructure image set through image segmentation based on the alloy microstructure image, and obtains a γ' phase microstructure image set through a denoising algorithm and image normalization. The alloy microstructure image after image segmentation and preprocessing can provide a high-precision modeling data set for the construction of an alloy microstructure recognition model.

[0078] In one embodiment, Figure 3 As shown, the γ' phase microstructure image set is input into the preset convolutional neural network model for training, and the convolutional neural network model for obtaining the alloy microstructure image includes:

[0079] S420, input the γ' phase microstructure image set to the input layer of the preset convolutional neural network model, and obtain feature data of the γ' phase microstructure image set through convolution and pooling operations in the hidden layer of the preset convolutional neural network model.

[0080] S440, according to the characteristic data, through the forward propagation function, obtain the stress and heat treatment state test results corresponding to the characteristic data, and send the test results to the output layer of the preset convolutional neural network model.

[0081] S460, according to the stress and heat treatment state test results, the parameters of the forward propagation function are updated through back propagation in the output layer to obtain the convolutional neural network model of the alloy microstructure image.

[0082] In this embodiment, the γ' phase microstructure image set is input to the input layer of the preset convolutional neural network model, and the characteristic data of the γ' phase microstructure image set is obtained by convolution and pooling operations in the hidden layer of the preset convolutional neural network model. In order to better represent the characteristics of the γ' phase, the characteristic data of the γ' phase microstructure image set includes the size, area, shape and other characteristic data that can characterize the characteristics of the γ' phase. The characteristic data of the γ' phase microstructure image set is one-to-one corresponding to the stress and heat treatment state. According to the characteristic data of the γ' phase microstructure image set, the stress and heat treatment state test results corresponding to the characteristic data are obtained through the forward propagation function, and the obtained test results are sent to the output layer of the preset convolutional neural network model. In the output layer, according to the stress and heat treatment state test results, the test results are reversely transmitted back to the hidden layer through back propagation, and the parameters of the forward propagation function are updated in the hidden layer. After multiple rounds of training, the final forward propagation function parameters are obtained. According to the final forward propagation function parameters, the convolutional neural network model of the alloy microstructure image is obtained.

[0083] The scheme of the above embodiment is to input the γ' phase microstructure image set into the input layer of the preset convolutional neural network model, obtain the feature data of the γ' phase microstructure image set through convolution and pooling operations in the hidden layer of the preset convolutional neural network model, obtain the stress and heat treatment state test results corresponding to the feature data through the forward propagation function, and send the test results to the output layer of the preset convolutional neural network model, update the parameters of the forward propagation function through back propagation, obtain the final forward propagation function parameters after multiple rounds of training, and construct a convolutional neural network model of the alloy microstructure image. By inputting the γ' phase microstructure image set into the preset convolutional neural network model, the convolutional neural network model of the alloy microstructure image is constructed, which can improve the efficiency of alloy microstructure recognition.

[0084] In one embodiment, Figure 4 As shown, the γ' phase microstructure image set is input into the input layer of the preset convolutional neural network model, and the feature data of the γ' phase microstructure image set is obtained through convolution and pooling operations in the hidden layer of the preset convolutional neural network model, including:

[0085] S421, input the γ' phase microstructure image set to the input layer of the preset convolutional neural network model, obtain the activated image set through the activation function, and pass the activated image set to the hidden layer of the preset convolutional neural network model.

[0086] S422, according to the activated image set, a feature vector is obtained through a convolution kernel and an activation function in a hidden layer.

[0087] S423, according to the feature vector, through the activation function and the pooling operation, obtain the pooled feature vector, and accumulate the pooling operation count value.

[0088] S424, after activating the pooled feature vector through an activation function, the activated image set is updated, and the step of obtaining the feature vector through the convolution kernel and the activation function in the hidden layer according to the activated image set is returned.

[0089] S425, if the accumulated pooling operation count value is greater than the preset value, characteristic data of the γ' phase microstructure image set is obtained according to the pooled characteristic vector.

[0090] In this embodiment, the hidden layer includes a convolution layer, a pooling layer, and a fully connected layer. The γ' phase microstructure image set is input into the input layer of the preset convolutional neural network model. Through the activation function, the activated image set is obtained. The activated image set is passed to the convolution layer of the preset convolutional neural network model via the connection weights between the neural networks. After the activated image set is subjected to the convolution operation of the convolution kernel and the activation function in the convolution layer, a feature vector is obtained. After the feature vector is added with a preset weight value, it is activated by the activation function and passed to the pooling layer. The pooling operation is performed in the pooling layer to obtain the pooled feature vector. After each pooling operation, the pooling operation count value is accumulated, wherein the pooling operation includes but is not limited to maximum pooling or average pooling. The pooled feature vector is then activated by the activation function and passed to the next layer of neurons for multi-layer convolution and pooling operations. Specifically, after the pooled feature vector is activated by the activation function, the activated image set is updated, and the step of obtaining the feature vector through the convolution kernel and the activation function in the hidden layer according to the activated image set is returned. If the pooling operation count value is greater than the pre-designed value, the pooled feature vector is obtained after multi-layer convolution kernel pooling operation, and the feature data of the γ' phase microstructure image set is obtained.

[0091] The scheme of the above-mentioned embodiment obtains the feature data of the γ' phase microstructure image set by inputting the γ' phase microstructure image set into the input layer of the preset convolutional neural network model through multi-layer convolution and pooling operations. The feature data of the γ' phase microstructure image set corresponds one-to-one to the alloy microstructure stress and heat treatment state, and can support obtaining the linear classification results of the stress and heat treatment state corresponding to the alloy microstructure image.

[0092] In one embodiment, Figure 5 As shown in the figure, according to the stress and heat treatment state test results, the parameters of the forward propagation function are updated by back propagation in the output layer, and the convolutional neural network model for obtaining the alloy microstructure image includes:

[0093] S462, according to the γ' phase microstructure image set, obtaining initial stress and heat treatment state data corresponding to the γ' phase microstructure image set.

[0094] S464, obtaining a loss function according to the stress and heat treatment state test results and the initial stress and heat treatment state data.

[0095] S466, if the loss function is greater than a preset threshold, the output layer updates the parameters of the forward propagation function by back propagating and calculating the gradient.

[0096] S468, if the loss function is less than or equal to the preset threshold, the feature data of the γ' phase microstructure image set is classified through a linear activation function in the output layer, the linear classification results of stress and heat treatment state corresponding to the alloy microstructure image are output, and a convolutional neural network model of the alloy microstructure image is constructed.

[0097] In this embodiment, according to the obtained γ' phase microstructure image set of different stresses and heat treatment states, the initial stress and heat treatment state data corresponding to the γ' phase microstructure image set are obtained by comparing the state data of the alloy microstructure under the standard stress and heat treatment state, and the stress and heat treatment state test results are subtracted from the initial stress and heat treatment state data to obtain a loss function. If the loss function is greater than a preset threshold, the output layer transmits the test results back to the hidden layer through back propagation to start a new round of training. In each round of training, the parameters of the forward propagation function are updated by calculating the gradient. After multiple rounds of training, if the loss function is less than or equal to the preset threshold, the training is stopped to obtain the final parameters of the forward propagation function. The feature data of the γ' phase microstructure image set are classified through a linear activation function in the output layer, and the stress and heat treatment state linear classification results corresponding to the alloy microstructure image are output. According to the parameters of the final forward propagation function, a convolutional neural network model of the alloy microstructure image is constructed.

[0098] The scheme of the above embodiment obtains a loss function according to the stress and heat treatment state test results, and by judging the loss function, starts multiple rounds of training through back propagation in the output layer, updates the parameters of the forward propagation function, and obtains a convolutional neural network model of the alloy microstructure image, which can support the linear classification results of the alloy microstructure. The feature data of the γ' phase microstructure image set is classified through a linear activation function in the output layer, and the linear classification results of the stress and heat treatment state corresponding to the alloy microstructure image are output. Since the feature data of the γ' phase microstructure image set corresponds to the alloy microstructure stress and heat treatment state one by one, the introduction of the linear activation function can output the linear classification results of the alloy microstructure stress and heat treatment state. Compared with the traditional logistic regression classification method, the linear activation function is used for classification, and any number of linear classification results of the alloy microstructure stress and heat treatment state can be obtained through any number of feature data types of the γ' phase microstructure image set, which is conducive to improving the classification accuracy and further improving the accuracy of the alloy microstructure recognition result.

[0099] In order to explain in detail the alloy microstructure identification model construction method and effect in this scheme, a most detailed embodiment is described below, wherein the alloy is a nickel-based single crystal high-temperature alloy:

[0100] The middle of the cross dendrite of the cross section of the nickel-based single crystal high-temperature alloy was selected as the observation area. Experiments were carried out for 500 hours at different stresses and temperatures to obtain alloy microstructure images of different stresses and heat treatment states. The 12 groups of temperature and stress selected are shown in Table 1, where one group of data includes one temperature data and one stress data:

[0101] Table 1 shows 12 selected sets of temperature and stress data

[0102] Temperature / ℃ Stress / MPa Temperature / ℃ Stress / MPa Temperature / ℃ Stress / MPa 1020 150 980 180 880 220 1020 120 980 140 880 160 1020 100 980 110 880 130 1020 70 980 70 880 70

[0103] Ten different areas are randomly selected on the test samples of the alloy under each stress and heat treatment state. These areas can clearly distinguish the γ matrix and γ' precipitation phase of the alloy microstructure under a magnification of 20,000 times. For the microstructure images of the alloy under all stress and heat treatment states, a SEM (Scanning Electron Microscope) is used to collect one SEM photo in each of the 10 areas, and the number of γ' phase particles in a single photo is not less than 50, so as to obtain an initial γ' phase microstructure image set. The initial γ' phase microstructure image set is denoised by median filtering, and then the image is normalized to obtain a γ' phase microstructure image set; the γ' phase microstructure image set is input into the input layer of the preset convolutional neural network model, and the activated image set is obtained through the activation function, and the activated image set is passed to the hidden layer of the preset convolutional neural network model. According to the activated image set, the feature vector is obtained in the hidden layer through the convolution kernel and the activation function, and the feature vector is obtained according to the feature vector. vector, obtain the pooled feature vector through the activation function and pooling operation, and accumulate the pooling operation count value, activate the pooled feature vector through the activation function, update the activated image set, and return to the step of obtaining the feature vector through the convolution kernel and the activation function in the hidden layer according to the activated image set. If the accumulated pooling operation count value is greater than the pre-designed value, obtain the feature data of the γ' phase microstructure image set according to the pooled feature vector, obtain the stress and heat treatment state test results corresponding to the feature data through the forward propagation function according to the feature data, and send the test results to the output layer of the preset convolutional neural network model, wherein the forward propagation function is: y=σ(w T x+b), w and b are random initialization parameters; σ=max(0,z) is the activation function. According to the γ' phase microstructure image set, the initial stress and heat treatment state data corresponding to the γ' phase microstructure image set are obtained. According to the stress and heat treatment state test results and the initial stress and heat treatment state data, the loss function is obtained. The expression of the loss function is: in a is the initial stress and heat treatment state data, and y is the test result obtained by the forward propagation function If the loss function is greater than a preset threshold, the output layer updates the parameters of the forward propagation function by back propagation and calculating the gradient. If the loss function is less than or equal to the preset threshold, the output layer uses a linear activation function to classify the feature data of the γ' phase microstructure image set, outputs the linear classification results of the stress and heat treatment state corresponding to the alloy microstructure image, and constructs a convolutional neural network model of the alloy microstructure image. The training process includes extracting the feature data of the γ' phase microstructure image set, classifying the feature data of the γ' phase microstructure image set through a linear activation function, and outputting the linear classification results of the stress and heat treatment state corresponding to the γ' phase microstructure image set.

[0104] In the above-mentioned alloy microstructure recognition model construction method, the alloy microstructure images of different stresses and heat treatment states are input into a preset convolutional neural network model for training after image segmentation to obtain a convolutional neural network model of the alloy microstructure image, which can improve the alloy microstructure recognition efficiency. Since the feature data of the γ' phase microstructure image set corresponds one-to-one to the stress and heat treatment state of the alloy microstructure image, the feature data of the γ' phase microstructure image set is classified through a linear activation function during the training process, and the linear classification results of the stress and heat treatment state corresponding to the γ' phase microstructure image set can be obtained.

[0105] It should be understood that, although the various steps in the flowcharts involved in the above-mentioned embodiments are displayed in sequence according to the indication of the arrows, these steps are not necessarily executed in sequence according to the order indicated by the arrows. Unless there is a clear explanation in this article, the execution of these steps does not have a strict order restriction, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above-mentioned embodiments can include multiple steps or multiple stages, and these steps or stages are not necessarily executed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a part of the steps or stages in other steps.

[0106] Based on the same inventive concept, the embodiment of the present application also provides an alloy microstructure identification model construction device for implementing the alloy microstructure identification model construction method involved above. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme recorded in the above method, so the specific limitations in one or more alloy microstructure identification model construction device embodiments provided below can refer to the limitations of the alloy microstructure identification model construction method above, and will not be repeated here.

[0107] In one embodiment, Figure 6As shown, an alloy microstructure identification model construction device 100 is provided, comprising: an image data acquisition module 120, an image segmentation module 140 and a model training module 160, wherein:

[0108] Image data acquisition module 120 is used to acquire alloy microstructure images of different stress and heat treatment states.

[0109] The image segmentation module 140 is used to obtain a set of γ' phase microstructure images by performing image segmentation based on the alloy microstructure image.

[0110] The model training module 160 is used to input the γ' phase microstructure image set into a preset convolutional neural network model for training to obtain a convolutional neural network model of the alloy microstructure image, wherein the training process includes extracting feature data of the γ' phase microstructure image set, classifying the feature data of the γ' phase microstructure image set through a linear activation function, and outputting the linear classification results of stress and heat treatment state corresponding to the γ' phase microstructure image set.

[0111] In the above-mentioned alloy microstructure recognition model construction device, the alloy microstructure images of different stresses and heat treatment states are input into a preset convolutional neural network model for training after image segmentation to obtain a convolutional neural network model of the alloy microstructure image, which can improve the alloy microstructure recognition efficiency. Since the feature data of the γ' phase microstructure image set corresponds one-to-one to the stress and heat treatment state of the alloy microstructure image, the feature data of the γ' phase microstructure image set is classified through a linear activation function during the training process, and the linear classification results of the stress and heat treatment state corresponding to the γ' phase microstructure image set can be obtained.

[0112] In one embodiment, the image segmentation module 140 is also used to obtain an initial γ' phase microstructure image set through image segmentation based on the alloy microstructure image; the initial γ' phase microstructure image set is subjected to a preprocessing method to obtain a γ' phase microstructure image set, and the preprocessing method includes a denoising algorithm and image normalization.

[0113] In one embodiment, the model training module 160 is also used to input the γ' phase microstructure image set into the input layer of the preset convolutional neural network model, and obtain the feature data of the γ' phase microstructure image set through convolution and pooling operations in the hidden layer of the preset convolutional neural network model; based on the feature data, the stress and heat treatment state test results corresponding to the feature data are obtained through the forward propagation function, and the test results are sent to the output layer of the preset convolutional neural network model; based on the stress and heat treatment state test results, the parameters of the forward propagation function are updated through back propagation in the output layer to obtain the convolutional neural network model of the alloy microstructure image.

[0114] In one embodiment, the model training module 160 is also used to input the γ' phase microstructure image set into the input layer of the preset convolutional neural network model, obtain the activated image set through the activation function, and pass the activated image set to the hidden layer of the preset convolutional neural network model; obtain the feature vector through the convolution kernel and the activation function in the hidden layer according to the activated image set; obtain the pooled feature vector through the activation function and the pooling operation according to the feature vector, and accumulate the pooling operation count value; after activating the pooled feature vector through the activation function, update the activated image set, and return to the step of obtaining the feature vector through the convolution kernel and the activation function in the hidden layer according to the activated image set; if the accumulated pooling operation count value is greater than the preset value, obtain the feature data of the γ' phase microstructure image set according to the pooled feature vector.

[0115] In one embodiment, the model training module 160 is also used to obtain the initial stress and heat treatment state data corresponding to the γ' phase microstructure image set based on the γ' phase microstructure image set; obtain the loss function based on the stress and heat treatment state test results and the initial stress and heat treatment state data; if the loss function is greater than a preset threshold, the output layer updates the parameters of the forward propagation function through back propagation and gradient calculation; if the loss function is less than or equal to the preset threshold, the output layer uses a linear activation function to classify the feature data of the γ' phase microstructure image set, outputs the linear classification results of the stress and heat treatment state corresponding to the alloy microstructure image, and constructs a convolutional neural network model of the alloy microstructure image.

[0116] Each module in the alloy microstructure identification model construction device can be implemented in whole or in part by software, hardware, or a combination thereof. Each module can be embedded in or independent of a processor in a computer device in the form of hardware, or can be stored in a memory in a computer device in the form of software, so that the processor can call and execute operations corresponding to each module.

[0117] In one embodiment, Figure 7 As shown, a method for identifying alloy microstructure is provided, and the method is applied to Figure 1 The terminal 102 in the example is used as an example to illustrate, and the following steps are included:

[0118] S500: obtaining a microstructure image of the alloy to be tested.

[0119] Among them, the microstructure image of the alloy to be tested corresponds to the alloy microstructure images under different stress and heat treatment states used to construct the alloy microstructure identification model. It is the microstructure image of the alloy to be tested selected in the actual test, and the microstructure image of the alloy to be tested corresponds to one of the stress and heat treatment states.

[0120] Specifically, a microstructure image of the alloy to be tested is obtained.

[0121] S600, obtaining a set of γ' phase microstructure images to be tested by image segmentation according to the microstructure image of the alloy to be tested.

[0122] The image of the alloy microstructure to be tested is subjected to the same image segmentation method as in the method for building the alloy microstructure recognition model, so as to obtain a set of images of the γ' phase microstructure to be tested.

[0123] Specifically, according to the microstructure image of the alloy to be tested, a set of γ' phase microstructure images to be tested is obtained through image segmentation, and the obtained set of γ' phase microstructure images to be tested is used as the input of the trained convolutional neural network model to obtain the alloy microstructure recognition result.

[0124] S700, inputting the γ' phase microstructure image set to be tested into the convolutional neural network model of the alloy microstructure image, and obtaining the linear classification results of stress and heat treatment state corresponding to the alloy microstructure image to be tested; wherein the convolutional neural network model of the alloy microstructure image is trained using the alloy microstructure recognition model construction method.

[0125] Among them, the γ' phase microstructure image set to be tested is input into the convolutional neural network model of the alloy microstructure image to obtain the linear classification results of stress and heat treatment state corresponding to the alloy microstructure image to be tested. The convolutional neural network model of the alloy microstructure image is trained using the alloy microstructure recognition model construction method.

[0126] Specifically, the γ' phase microstructure image set to be tested is input into the convolutional neural network model of the alloy microstructure image to obtain the linear classification results of stress and heat treatment state corresponding to the alloy microstructure image to be tested. This method of using the convolutional neural network model to identify the alloy microstructure improves the efficiency of alloy microstructure identification.

[0127] In the above alloy microstructure identification method, the stress and heat treatment state linear classification results corresponding to the alloy microstructure image to be tested are obtained by inputting the image into the convolutional neural network model of the alloy microstructure image after image segmentation, wherein the convolutional neural network model of the alloy microstructure image is trained using the alloy microstructure identification model construction method. This method of alloy microstructure identification using the convolutional neural network model improves the efficiency of alloy microstructure identification.

[0128] In order to explain the alloy microstructure identification method and effect in this scheme in detail, a most detailed embodiment is described below, wherein the alloy to be tested is a nickel-based single crystal high-temperature alloy, the stress is 100MPa, and the temperature corresponding to the heat treatment state is 1020°C:

[0129] The middle of the cross dendrite of the cross section of the nickel-based single crystal high-temperature alloy was selected as the observation area, and the microstructure image of the alloy to be tested at 100MPa and 1020℃ was obtained. Ten different areas were randomly selected on the test sample of the alloy. These areas can clearly distinguish the γ matrix and γ' precipitation phase of the alloy microstructure at a magnification of 20,000 times. One SEM photo was collected in each of the 10 areas, and the number of γ' phase particles in a single photo was not less than 50, and the initial γ' phase microstructure image set to be tested was obtained. The initial γ' phase microstructure image set to be tested was filtered through a median filter. The image set of γ' phase microstructure to be tested is obtained by denoising and then image normalization; the image set of γ' phase microstructure to be tested is input into the convolutional neural network model of alloy microstructure image, and the linear classification results of stress and heat treatment state corresponding to the alloy microstructure image to be tested are obtained. The classification results show that the stress and heat treatment state corresponding to the alloy microstructure image to be tested are 100MPa and 1020℃, respectively, which are consistent with the stress and heat treatment state corresponding to the alloy to be tested. Among them, the convolutional neural network model of alloy microstructure image is trained by alloy microstructure recognition model construction method.

[0130] The above-mentioned alloy microstructure identification method, device, computer equipment and storage medium obtain the linear classification results of stress and heat treatment state corresponding to the alloy microstructure image to be tested by inputting the image into the convolutional neural network model of the alloy microstructure image after image segmentation, wherein the convolutional neural network model of the alloy microstructure image is trained using the alloy microstructure identification model construction method. This method of alloy microstructure identification using the convolutional neural network model improves the efficiency of alloy microstructure identification.

[0131] It should be understood that, although the various steps in the flowcharts involved in the above-mentioned embodiments are displayed in sequence according to the indication of the arrows, these steps are not necessarily executed in sequence according to the order indicated by the arrows. Unless there is a clear explanation in this article, the execution of these steps does not have a strict order restriction, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above-mentioned embodiments can include multiple steps or multiple stages, and these steps or stages are not necessarily executed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a part of the steps or stages in other steps.

[0132] Based on the same inventive concept, the embodiment of the present application also provides an alloy microstructure identification device for implementing the alloy microstructure identification method involved above. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme recorded in the above method, so the specific limitations in one or more alloy microstructure identification device embodiments provided below can refer to the limitations of the alloy microstructure identification method above, and will not be repeated here.

[0133] In one embodiment, Figure 8 As shown, an alloy microstructure identification device 800 is provided, comprising: a test image data acquisition module 820, a test image segmentation module 840 and a classification result acquisition module 860, wherein:

[0134] The image data acquisition module 820 is used to obtain the microstructure image of the alloy to be tested.

[0135] The image segmentation module 840 is used to obtain a set of γ' phase microstructure images to be tested by image segmentation according to the microstructure image of the alloy to be tested.

[0136] The classification result acquisition module 860 is used to input the γ' phase microstructure image set to be tested into the convolutional neural network model of the alloy microstructure image to obtain the linear classification results of stress and heat treatment state corresponding to the alloy microstructure image to be tested; wherein, the convolutional neural network model of the alloy microstructure image is trained using the alloy microstructure recognition model construction method.

[0137] In the above-mentioned alloy microstructure identification device, the linear classification results of stress and heat treatment state corresponding to the alloy microstructure image to be tested are obtained by inputting the convolutional neural network model into the alloy microstructure image after image segmentation, wherein the convolutional neural network model of the alloy microstructure image is trained using the alloy microstructure identification model construction method. This method of alloy microstructure identification using a convolutional neural network model improves the efficiency of alloy microstructure identification.

[0138] Each module in the alloy microstructure identification device can be implemented in whole or in part by software, hardware, or a combination thereof. Each module can be embedded in or independent of a processor in a computer device in the form of hardware, or can be stored in a memory in a computer device in the form of software, so that the processor can call and execute operations corresponding to each module.

[0139] In one embodiment, a computer device is provided. The computer device may be a server, and its internal structure diagram may be as follows: Fig. 9As shown. The computer device includes a processor, a memory and a network interface connected through a system bus. 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 a convolutional neural network model of the alloy microstructure image and the linear classification results of the stress and heat treatment state corresponding to the alloy microstructure image. The network 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, an alloy microstructure identification method is implemented.

[0140] 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.

[0141] In one embodiment, a computer device is provided, including a memory and a processor, wherein a computer program is stored in the memory, and when the processor executes the computer program, the following steps are implemented:

[0142] Obtain microstructure images of alloys in different stress and heat treatment states;

[0143] According to the alloy microstructure image, a γ' phase microstructure image set is obtained by image segmentation;

[0144] The γ' phase microstructure image set is input into a preset convolutional neural network model for training to obtain a convolutional neural network model of the alloy microstructure image, wherein the training process includes extracting feature data of the γ' phase microstructure image set, classifying the feature data of the γ' phase microstructure image set through a linear activation function, and outputting the linear classification results of stress and heat treatment state corresponding to the γ' phase microstructure image set.

[0145] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored, and when the computer program is executed by a processor, the following steps are implemented:

[0146] Obtain microstructure images of alloys in different stress and heat treatment states;

[0147] According to the alloy microstructure image, a γ' phase microstructure image set is obtained by image segmentation;

[0148] The γ' phase microstructure image set is input into a preset convolutional neural network model for training to obtain a convolutional neural network model of the alloy microstructure image, wherein the training process includes extracting feature data of the γ' phase microstructure image set, classifying the feature data of the γ' phase microstructure image set through a linear activation function, and outputting the linear classification results of stress and heat treatment state corresponding to the γ' phase microstructure image set.

[0149] In one embodiment, a computer program product is provided, comprising a computer program, which, when executed by a processor, implements the following steps:

[0150] Obtain microstructure images of alloys in different stress and heat treatment states;

[0151] According to the alloy microstructure image, a γ' phase microstructure image set is obtained by image segmentation;

[0152] The γ' phase microstructure image set is input into a preset convolutional neural network model for training to obtain a convolutional neural network model of the alloy microstructure image, wherein the training process includes extracting feature data of the γ' phase microstructure image set, classifying the feature data of the γ' phase microstructure image set through a linear activation function, and outputting the linear classification results of stress and heat treatment state corresponding to the γ' phase microstructure image set.

[0153] 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.

[0154] Those skilled 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), magnetoresistive 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. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. Non-relational databases may include distributed databases based on blockchains, etc., but are not limited to this. 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 are not limited to this.

[0155] 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.

[0156] The above-described embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present application. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.

Claims

1. A method for constructing an alloy microstructure identification model, characterized in that: The method comprises: Obtain microstructure images of alloys in different stress and heat treatment states; According to the alloy microstructure image, γ ’ Phase microstructure image set; The γ ’ The phase microstructure image set is input into a preset convolutional neural network model for training to obtain a convolutional neural network model of the alloy microstructure image, wherein the training process includes extracting the γ ’ If the loss function is less than or equal to the preset threshold, the γ ’ The characteristic data of the phase microstructure image set is classified and the γ ’ The linear classification results of stress and heat treatment state corresponding to the phase microstructure image set; the characteristic data corresponds to the stress and heat treatment state of the alloy microstructure image one by one; the loss function is obtained according to the stress and heat treatment state test results and the initial stress and heat treatment state data; the initial stress and heat treatment state data are obtained according to the γ ’ The phase microstructure image set is obtained; the stress and heat treatment state test results are obtained based on the characteristic data through the forward propagation function.

2. The method for constructing an alloy microstructure identification model according to claim 1, characterized in that: According to the alloy microstructure image, the γ ’ Phase microstructure image sets include: According to the alloy microstructure image, the initial γ ’ Phase microstructure image set; The initial γ ’ Phase microstructure image set, through preprocessing method, obtain γ ’ Phase microstructure image set, the preprocessing method includes a denoising algorithm and image normalization.

3. The method for constructing an alloy microstructure identification model according to claim 1, characterized in that: The γ ’ The phase microstructure image set is input into a preset convolutional neural network model for training, and the convolutional neural network model for obtaining the alloy microstructure image includes: The γ ’ The phase microstructure image set is input into the input layer of the preset convolutional neural network model, and the γ ’ Characteristic data of phase microstructure image set; According to the characteristic data, obtaining stress and heat treatment state test results corresponding to the characteristic data through a forward propagation function, and sending the test results to the output layer of the preset convolutional neural network model; According to the stress and heat treatment state test results, the parameters of the forward propagation function are updated through back propagation in the output layer to obtain a convolutional neural network model of the alloy microstructure image.

4. The method for constructing an alloy microstructure identification model according to claim 3, characterized in that: The γ ’ The phase microstructure image set is input into the input layer of the preset convolutional neural network model, and the γ ’ The characteristic data of the phase microstructure image set include: The γ ’ The phase microstructure image set is input into the input layer of the preset convolutional neural network model, an activated image set is obtained through an activation function, and the activated image set is passed to the hidden layer of the preset convolutional neural network model; According to the activated image set, a feature vector is obtained by using a convolution kernel and an activation function in the hidden layer; According to the feature vector, through an activation function and a pooling operation, a pooled feature vector is obtained, and a count value of the pooling operation is accumulated; After activating the pooled feature vector through an activation function, updating the activated image set, and returning the step of obtaining the feature vector through the convolution kernel and the activation function in the hidden layer according to the activated image set; If the pooling operation count value is greater than the pre-designed value, the γ is obtained according to the feature vector after pooling. ’ Characteristic data of phase microstructure image set.

5. The method for constructing an alloy microstructure identification model according to claim 3, characterized in that: The convolutional neural network model for obtaining the alloy microstructure image by updating the parameters of the forward propagation function in the output layer through back propagation according to the stress and heat treatment state test results includes: According to the ’ Phase microstructure image set, obtain the γ ’ Initial stress and heat treatment state data corresponding to the phase microstructure image set; Obtaining a loss function according to the stress and heat treatment state test results and the initial stress and heat treatment state data; If the loss function is greater than a preset threshold, the output layer updates the parameters of the forward propagation function by back propagating and calculating the gradient; If the loss function is less than or equal to the preset threshold, the γ ’ The characteristic data of the phase microstructure image set are classified, the linear classification results of stress and heat treatment state corresponding to the alloy microstructure image are output, and a convolutional neural network model of the alloy microstructure image is constructed.

6. A method for identifying alloy microstructure, characterized in that: The method comprises: Obtaining microstructure images of the alloy to be tested; According to the microstructure image of the alloy to be tested, the γ ’ Phase microstructure image set; The γ ’ The phase microstructure image set is input into the convolutional neural network model of the alloy microstructure image to obtain the linear classification results of stress and heat treatment state corresponding to the alloy microstructure image to be tested; wherein the convolutional neural network model of the alloy microstructure image is trained using the alloy microstructure recognition model construction method according to any one of claims 1 to 5.

7. A device for constructing an alloy microstructure identification model, characterized in that: The device comprises: Image data acquisition module, used to obtain alloy microstructure images under different stress and heat treatment states; The image segmentation module is used to obtain γ ’ Phase microstructure image set; Model training module, used to transform the γ ’ The phase microstructure image set is input into a preset convolutional neural network model for training to obtain a convolutional neural network model of the alloy microstructure image, wherein the training process includes extracting the γ ’ If the loss function is less than or equal to the preset threshold, the γ ’ The characteristic data of the phase microstructure image set is classified and the γ ’ The linear classification results of stress and heat treatment state corresponding to the phase microstructure image set; the characteristic data corresponds to the stress and heat treatment state of the alloy microstructure image one by one; the loss function is obtained according to the stress and heat treatment state test results and the initial stress and heat treatment state data; the initial stress and heat treatment state data are obtained according to the γ ’ The phase microstructure image set is obtained; the stress and heat treatment state test results are obtained based on the characteristic data through the forward propagation function.

8. An alloy microstructure identification device, characterized in that: The device comprises: The image data acquisition module to be tested is used to obtain the microstructure image of the alloy to be tested; The image segmentation module is used to obtain the γ ’ Phase microstructure image set; The classification result acquisition module is used to obtain the γ ’ The phase microstructure image set is input into the convolutional neural network model of the alloy microstructure image to obtain the linear classification results of stress and heat treatment state corresponding to the alloy microstructure image to be tested; wherein the convolutional neural network model of the alloy microstructure image is trained using the alloy microstructure recognition model construction method according to any one of claims 1 to 5.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the processor implements the steps of the alloy microstructure identification model construction method described in any one of claims 1 to 5 or the alloy microstructure identification method described in claim 6.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the alloy microstructure identification model construction method described in any one of claims 1 to 5 or the alloy microstructure identification method described in claim 6 are implemented.

Images