Alloy sensitization degree analysis method based on metallographic diagram
By introducing the frequency channel attention mechanism and frequency domain analysis into the ResNet model, the problems of low efficiency and insufficient accuracy in traditional methods are solved, and efficient and accurate alloy sensitization analysis is achieved.
Patent Information
- Application Number
- CN202510683654.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-05-26
- Publication Date
- 2025-09-05
AI Technical Summary
Existing alloy sensitization analysis methods based on metallographic images have problems such as low analysis efficiency, insufficient accuracy and poor adaptability. Especially when processing complex metallographic images, the traditional ResNet model cannot fully utilize the frequency domain information.
The FcaNet model is constructed using the frequency channel attention mechanism based on the ResNet structure to preprocess and extract features of metallographic images. The frequency components are extracted through Fourier transform or discrete cosine transform, and the alloy sensitization is classified in combination with the Softmax function.
It improves the efficiency, accuracy and adaptability of alloy sensitization analysis, realizes high-precision quantitative analysis, and enhances the feature extraction and classification capabilities of metallographic images.
Smart Images
Figure CN120599339A_ABST
Abstract
Description
Technical Field
[0001] The embodiments of the present disclosure relate to the field of measurement technology, and in particular to an alloy sensitization analysis method based on a metallographic diagram. Background Art
[0002] At present, the sensitization degree of alloy materials is an important indicator to measure their corrosion resistance under specific environments. The sensitization process usually leads to the formation of precipitation phases at the grain boundaries of alloy materials, thereby reducing their corrosion resistance. Metallographic analysis, as an important means of microstructural characterization, can effectively reveal information such as the grain size, shape, and phase composition of alloy materials, thereby providing an important basis for the sensitivity analysis of materials. However, traditional metallographic sensitivity analysis methods mostly rely on manual interpretation and empirical judgment, which is not only inefficient but also easily affected by human factors, resulting in inconsistency and lack of accuracy in the analysis results.
[0003] Automated microstructural characterization methods based on image analysis utilize deep learning models such as convolutional neural networks (CNNs) to automatically analyze and classify metallographic images. This not only improves the speed and accuracy of analysis, but also enables the extraction of more refined features from large amounts of data, revealing microstructural information that is difficult to detect using traditional methods. However, existing deep learning methods often suffer from insufficient model generalization and inadequate feature extraction when processing complex metallographic images.
[0004] As an effective image processing algorithm, deep residual neural networks (ResNet) are widely used in various image classification and recognition tasks. However, when processing the microstructure of metallographic images, traditional ResNet often fails to fully utilize the frequency domain information in the image, thus affecting the precision and accuracy of sensitization analysis.
[0005] It can be seen that there is an urgent need for an alloy sensitization analysis method based on metallographic diagrams with high analysis efficiency, accuracy and adaptability. Summary of the Invention
[0006] In view of this, the embodiments of the present disclosure provide an alloy sensitization analysis method based on metallographic diagrams, which at least partially solves the problems of poor analysis efficiency, accuracy and adaptability in the prior art.
[0007] The present disclosure provides a method for analyzing alloy sensitization based on a metallographic diagram, comprising:
[0008] Step 1: After pre-processing different sample alloys, a preset number of metallographic images of each sample alloy are collected to form a sample data set;
[0009] Step 2: After preprocessing the sample data set, divide it into training set, validation set and test set;
[0010] Step 3: Based on the ResNet structure, the frequency channel attention mechanism is introduced to build the FcaNet model;
[0011] Step 4: Use the training set and validation set to train the FcaNet model to obtain the sensitization analysis model, and then use the test set to input the sensitization analysis model to establish a quantitative standard;
[0012] Step 5: Collect the metallographic image of the target alloy and input it into the sensitization analysis model to obtain the probability distribution, and combine the probability distribution and the quantitative standard to obtain the sensitization classification result of the target alloy.
[0013] According to a specific implementation of the embodiment of the present disclosure, step 1 specifically includes:
[0014] Step 1.1: Cut different sample alloys into preset sizes using mechanical methods and perform cold mounting.
[0015] Step 1.2, grinding and polishing the sample alloy;
[0016] Step 1.3, cleaning the mounted and polished sample alloy and then performing corrosion treatment;
[0017] Step 1.4: Under consistent lighting conditions, collect a preset number of metallographic images of each sample alloy after corrosion treatment and form a sample data set.
[0018] According to a specific implementation of the embodiment of the present disclosure, step 2 specifically includes:
[0019] Step 2.1, convert the color metallographic image into a grayscale image;
[0020] Step 2.2, denoising the grayscale image;
[0021] In step 2.3, the pixel values of the denoised grayscale image are normalized and divided into training set, validation set and test set.
[0022] According to a specific implementation of the embodiment of the present disclosure, the FcaNet model includes:
[0023] A convolutional layer, which performs a convolution operation on the input image to extract underlying features;
[0024] A pooling layer that downsamples feature maps to reduce computational complexity and retain key information;
[0025] An activation function layer, which uses the ReLU activation function to introduce nonlinearity and enhance the model's expressiveness;
[0026] A fully connected layer maps the features extracted by the convolutional layer to a feature space;
[0027] A frequency channel attention mechanism is provided, which is used to enhance the feature extraction capability of metallographic images.
[0028] According to a specific implementation of the embodiment of the present disclosure, step 4 specifically includes:
[0029] Step 4.1: Input the training set into the FcaNet model to obtain the analysis results and calculate the cross entropy loss function based on them;
[0030] In step 4.2, based on the cross entropy loss function, the preset optimization algorithm is used to continuously update the FcaNet model parameters through back propagation to optimize the model performance.
[0031] Step 4.3: During the training process, the model performance is evaluated using the validation set until it meets the training requirements and the sensitivity analysis model is obtained;
[0032] In step 4.4, the test set is input into the sensitization analysis model, and the sensitization classification results of each sample alloy and their corresponding probability distribution are output, thereby forming a quantitative standard for sensitization classification.
[0033] According to a specific implementation of the embodiment of the present disclosure, step 5 specifically includes:
[0034] Step 5.1: Collect the metallographic image of the target alloy and input it into the sensitization analysis model. Use Fourier transform or discrete cosine transform in the channel attention mechanism to extract the frequency components of the metallographic image.
[0035] Step 5.2: Analyze the frequency components from low frequency to high frequency components step by step to extract multi-level features;
[0036] In step 5.3, the multi-level features are classified by the Softmax function, the probability distribution of the alloy sensitization is output, and the sensitization classification result of the target alloy is obtained by combining the quantitative standard.
[0037] The alloy sensitization analysis scheme based on metallographic images in the embodiment of the present disclosure includes: step 1, pre-processing different sample alloys and collecting a preset number of metallographic images of each sample alloy to form a sample data set; step 2, pre-processing the sample data set and dividing it into a training set, a validation set and a test set; step 3, introducing a frequency channel attention mechanism based on the ResNet structure to construct an FcaNet model; step 4, using the training set and the validation set to train the FcaNet model to obtain a sensitization analysis model, and then using the test set to input the sensitization analysis model to establish a quantitative standard; step 5, collecting the metallographic image of the target alloy and inputting it into the sensitization analysis model to obtain a probability distribution, and combining the probability distribution and the quantitative standard to obtain the sensitization classification result of the target alloy.
[0038] The beneficial effects of the embodiments of the present disclosure are as follows: through the scheme of the present disclosure, channel attention is rethought through frequency domain analysis on the basis of the traditional Resnet algorithm, the feature extraction and classification capabilities of metallographic images are further improved, high-precision quantitative analysis of alloy sensitization is achieved, and analysis efficiency, accuracy and adaptability are improved. BRIEF DESCRIPTION OF THE DRAWINGS
[0039] In order to more clearly illustrate the technical solutions of the embodiments of the present disclosure, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present disclosure. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.
[0040] Figure 1 A schematic flow chart of a method for analyzing alloy sensitization based on a metallographic diagram provided in an embodiment of the present disclosure;
[0041] Figure 2 This is a flowchart of an algorithm for analyzing alloy sensitization based on metallographic diagrams according to an embodiment of the present disclosure. DETAILED DESCRIPTION
[0042] The embodiments of the present disclosure are described in detail below with reference to the accompanying drawings.
[0043] The following describes the embodiments of the present disclosure through specific concrete examples, and those skilled in the art can easily understand other advantages and effects of the present disclosure from the contents disclosed in this specification. Obviously, the described embodiments are only a part of the embodiments of the present disclosure, rather than all of the embodiments. The present disclosure can also be implemented or applied through other different specific embodiments, and the details in this specification can also be modified or changed in various ways based on different viewpoints and applications without departing from the spirit of the present disclosure. It should be noted that, in the absence of conflict, the following embodiments and features in the embodiments can be combined with each other. Based on the embodiments in the present disclosure, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present disclosure.
[0044] It should be noted that various aspects of the embodiments within the scope of the appended claims are described below. It should be apparent that the aspects described herein can be embodied in a wide variety of forms, and any specific structure and / or function described herein is merely illustrative. Based on this disclosure, it should be understood by those skilled in the art that an aspect described herein can be implemented independently of any other aspect, and two or more of these aspects can be combined in various ways. For example, any number of aspects described herein can be used to implement an apparatus and / or practice a method. In addition, other structures and / or functionalities other than one or more of the aspects described herein can be used to implement this apparatus and / or practice this method.
[0045] It should also be noted that the illustrations provided in the following embodiments are only schematic illustrations of the basic concept of the present disclosure. The illustrations only show components related to the present disclosure and are not drawn according to the number, shape and size of components in actual implementation. In actual implementation, the form, quantity and proportion of each component may be changed at will, and the component layout may also be more complicated.
[0046] Additionally, in the following description, specific details are provided to provide a thorough understanding of the examples. However, one skilled in the art will appreciate that the aspects described can be practiced without these specific details.
[0047] The present disclosure provides an alloy sensitization analysis method based on a metallographic diagram, which can be applied to the sensitization analysis process of alloy materials in industrial scenarios.
[0048] See also Figure 1 , is a flow chart of a method for analyzing alloy sensitization based on a metallographic diagram provided by an embodiment of the present disclosure. Figure 1 and Figure 2 As shown, the method mainly includes the following steps:
[0049] Step 1: After pre-processing different sample alloys, a preset number of metallographic images of each sample alloy are collected to form a sample data set;
[0050] Furthermore, the step 1 specifically includes:
[0051] Step 1.1: Cut different sample alloys into preset sizes using mechanical methods and perform cold mounting.
[0052] Step 1.2, grinding and polishing the sample alloy;
[0053] Step 1.3, cleaning the mounted and polished sample alloy and then performing corrosion treatment;
[0054] Step 1.4: Under consistent lighting conditions, collect a preset number of metallographic images of each sample alloy after corrosion treatment and form a sample data set.
[0055] In specific implementation, the metallographic sample preparation and image acquisition process can be as follows:
[0056] A. Mechanically cut the alloy sample into appropriate sizes and cold mount it with epoxy resin and amine curing agent to form a cylindrical sample with a diameter of 30 mm, containing 4-6 samples to avoid the effects of heating and pressurization on the sample;
[0057] B. Grind and polish the sample. To reduce interference and obtain an ideal metallographic image, use a grinder and polisher to perform steps such as coarse grinding, fine grinding, coarse polishing, and fine polishing in sequence until the sample surface reaches a mirror finish. Parameters such as time, pressure, and speed must be strictly controlled in each step to ensure sample quality.
[0058] C. After grinding and polishing, place the mounted and polished sample in an ultrasonic cleaner and clean it with anhydrous ethanol to remove surface impurities. Next, perform an etching treatment. For aluminum-magnesium alloy, for example, the etching reagent is Keller's reagent (1% HF + 1.5% HCl + 2.5% HNO3 + 95% H2O). The etching method is drop etching, and the etching time is based on the quality of the metallographic image.
[0059] D. Use a high-resolution microscope to capture images of the samples, ensuring that at least a certain number of metallographic images are collected for each sample. Consistent lighting conditions must be maintained during image acquisition to minimize errors.
[0060] Step 2: After preprocessing the sample data set, divide it into training set, validation set and test set;
[0061] Furthermore, the step 2 specifically includes:
[0062] Step 2.1, convert the color metallographic image into a grayscale image;
[0063] Step 2.2, denoising the grayscale image;
[0064] In step 2.3, the pixel values of the denoised grayscale image are normalized and divided into training set, validation set and test set.
[0065] In specific implementation, image preprocessing includes: grayscale conversion: converting color metallographic images into grayscale images to reduce information redundancy and retain structural features; denoising: using denoising methods such as Gaussian filtering or median filtering to remove noise and ensure image quality; normalization: normalizing image pixel values to the range of [0,1] to reduce the computational complexity caused by numerical differences and enhance the convergence of the model.
[0066] The preprocessed image samples are then divided into training set, validation set and test set in a ratio of 7:2:1 to ensure that each part can fully reflect the overall distribution of the samples. The training set, validation set and test set are used for subsequent training and feature extraction of the FcaNet network.
[0067] Step 3: Based on the ResNet structure, the frequency channel attention mechanism is introduced to build the FcaNet model;
[0068] Furthermore, the FcaNet model includes:
[0069] A convolutional layer, which performs a convolution operation on the input image to extract underlying features;
[0070] A pooling layer that downsamples feature maps to reduce computational complexity and retain key information;
[0071] An activation function layer, which uses the ReLU activation function to introduce nonlinearity and enhance the model's expressiveness;
[0072] A fully connected layer maps the features extracted by the convolutional layer to a feature space;
[0073] A frequency channel attention mechanism is provided, which is used to enhance the feature extraction capability of metallographic images.
[0074] The specific implementation process for building the FcaNet model involves building an FcaNet model based on the ResNet architecture and introducing a frequency channel attention mechanism to enhance feature extraction from metallographic images. The model structure primarily includes: a convolutional layer, which performs convolution operations on the input image to extract underlying features; a pooling layer, which downsamples the feature maps to reduce computational complexity while retaining key information; an activation layer, which uses the ReLU activation function to introduce nonlinearity and enhance the model's expressiveness; and a fully connected layer, which maps the features extracted by the convolutional layer to a feature space.
[0075] Step 4: Use the training set and validation set to train the FcaNet model to obtain the sensitization analysis model, and then use the test set to input the sensitization analysis model to establish a quantitative standard;
[0076] Based on the above embodiment, step 4 specifically includes:
[0077] Step 4.1: Input the training set into the FcaNet model to obtain the analysis results and calculate the cross entropy loss function based on them;
[0078] In step 4.2, based on the cross entropy loss function, the preset optimization algorithm is used to continuously update the FcaNet model parameters through back propagation to optimize the model performance.
[0079] Step 4.3: During the training process, the model performance is evaluated using the validation set until it meets the training requirements and the sensitivity analysis model is obtained;
[0080] In step 4.4, the test set is input into the sensitization analysis model, and the sensitization classification results of each sample alloy and their corresponding probability distribution are output, thereby forming a quantitative standard for sensitization classification.
[0081] In specific implementation, the model training and tuning process is as follows:
[0082] A. Model Training: The preprocessed training set is fed into the FcaNet network for training. The following strategies are used during training: Loss Function: A cross-entropy loss function is used to measure the difference between the model's analysis results and the actual labels; Optimization Algorithm: Adam or SGD optimization algorithms are used to continuously update network parameters through backpropagation to optimize model performance.
[0083] B. Model Tuning: During training, evaluate model performance using a validation set, monitoring metrics such as accuracy and loss. Tune the model by adjusting model parameters (such as learning rate and batch size) to prevent overfitting and ensure generalization.
[0084] C. Test sample classification: Use the trained FcaNet model to classify the test set samples and output the sensitivity classification results and corresponding probability distribution of each sample;
[0085] D. Quantitative Analysis: Based on the classification results, the sensitization degree of the alloy is quantitatively analyzed. By statistically analyzing the proportion of samples in different sensitization categories, a scientific quantitative description of the sensitization characteristics of the alloy material is provided;
[0086] E. Classification effect evaluation: To further verify the accuracy of model classification, indicators such as confusion matrix and ROC curve can be used to evaluate and verify the classification effect of the model to ensure that the classification results are sufficiently reliable and accurate.
[0087] Step 5: Collect the metallographic image of the target alloy and input it into the sensitization analysis model to obtain the probability distribution, and combine the probability distribution and the quantitative standard to obtain the sensitization classification result of the target alloy.
[0088] Based on the above embodiment, step 5 specifically includes:
[0089] Step 5.1: Collect the metallographic image of the target alloy and input it into the sensitization analysis model. Use Fourier transform or discrete cosine transform in the channel attention mechanism to extract the frequency components of the metallographic image.
[0090] Step 5.2: Analyze the frequency components from low frequency to high frequency components step by step to extract multi-level features;
[0091] In step 5.3, the multi-level features are classified by the Softmax function, the probability distribution of the alloy sensitization is output, and the sensitization classification result of the target alloy is obtained by combining the quantitative standard.
[0092] In specific implementation, a metallographic image of the target alloy is collected and input into the sensitization analysis model. Within the sensitization analysis model's channel attention mechanism, Fourier transform or discrete cosine transform is used to extract the frequency components of the metallographic image, enhancing the network's ability to capture global and local features. Multi-level feature extraction is performed using frequency domain information, with step-by-step analysis from low-frequency to high-frequency components. This further enhances the model's sensitivity to image details and global structure, ensuring accurate sensitization classification. The extracted image features are then classified using the Softmax function, outputting a probability distribution of the alloy's sensitization, ensuring that the sum of the probabilities of the classification results is 1.
[0093] The alloy sensitization analysis method based on metallographic images provided in this embodiment rethinks channel attention through frequency domain analysis on the basis of the traditional ResNet algorithm, further improves the feature extraction and classification capabilities of metallographic images, realizes high-precision quantitative analysis of alloy sensitization, and improves analysis efficiency, accuracy and adaptability.
[0094] It should be understood that various parts of the present disclosure can be implemented in hardware, software, firmware, or a combination thereof.
[0095] The above description is merely a specific embodiment of the present disclosure, but the scope of protection of the present disclosure is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this disclosure should be included in the scope of protection of the present disclosure. Therefore, the scope of protection of the present disclosure should be based on the scope of protection of the claims.
Claims
1. A method for analyzing alloy sensitization based on metallographic diagram, characterized in that: include: Step 1: After pre-processing different sample alloys, a preset number of metallographic images of each sample alloy are collected to form a sample data set; Step 2: After preprocessing the sample data set, divide it into training set, validation set and test set; Step 3: Based on the ResNet structure, the frequency channel attention mechanism is introduced to build the FcaNet model; Step 4: Use the training set and validation set to train the FcaNet model to obtain the sensitization analysis model, and then use the test set to input the sensitization analysis model to establish a quantitative standard; Step 5: Collect the metallographic image of the target alloy and input it into the sensitization analysis model to obtain the probability distribution, and combine the probability distribution and the quantitative standard to obtain the sensitization classification result of the target alloy.
2. The method according to claim 1, characterized in that , the step 1 specifically includes: Step 1.1: Cut different sample alloys into preset sizes using mechanical methods and perform cold mounting. Step 1.2, grinding and polishing the sample alloy; Step 1.3, cleaning the mounted and polished sample alloy and then performing corrosion treatment; Step 1.4: Under consistent lighting conditions, collect a preset number of metallographic images of each sample alloy after corrosion treatment and form a sample data set.
3. The method according to claim 2, characterized in that , the step 2 specifically includes: Step 2.1, convert the color metallographic image into a grayscale image; Step 2.2, denoising the grayscale image; In step 2.3, the pixel values of the denoised grayscale image are normalized and divided into training set, validation set and test set.
4. The method according to claim 3, characterized in that ,The FcaNet model includes: A convolutional layer, which performs a convolution operation on the input image to extract underlying features; A pooling layer that downsamples feature maps to reduce computational complexity and retain key information; An activation function layer, which uses the ReLU activation function to introduce nonlinearity and enhance the model's expressiveness; A fully connected layer maps the features extracted by the convolutional layer to a feature space; A frequency channel attention mechanism is provided, which is used to enhance the feature extraction capability of metallographic images.
5. The method according to claim 4, characterized in that , the step 4 specifically includes: Step 4.1: Input the training set into the FcaNet model to obtain the analysis results and calculate the cross entropy loss function based on them; In step 4.2, based on the cross entropy loss function, the preset optimization algorithm is used to continuously update the FcaNet model parameters through back propagation to optimize the model performance. Step 4.3: During the training process, the model performance is evaluated through the validation set until it meets the training requirements and the sensitivity analysis model is obtained; In step 4.4, the test set is input into the sensitization analysis model, and the sensitization classification results of each sample alloy and their corresponding probability distribution are output, thereby forming a quantitative standard for sensitization classification.
6. The method according to claim 5, characterized in that , the step 5 specifically includes: Step 5.1: Collect the metallographic image of the target alloy and input it into the sensitization analysis model. Use Fourier transform or discrete cosine transform in the channel attention mechanism to extract the frequency components of the metallographic image. Step 5.2: Analyze the frequency components from low frequency to high frequency components step by step to extract multi-level features; In step 5.3, the multi-level features are classified by the Softmax function, the probability distribution of the alloy sensitization is output, and the sensitization classification result of the target alloy is obtained by combining the quantitative standard.