Method for predicting fatigue crack size and stress intensity factor range of metal material

By conducting fatigue tests on metal materials and training of VGG-16 network model, the problem of difficult to distinguish between fatigue band spacing and low computational efficiency is solved, and more accurate and efficient quantitative prediction of fatigue fractures is achieved.

CN118918998APending Publication Date: 2024-11-08XI AN JIAOTONG UNIV

Patent Information

Application Number
CN202411041092.8
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-31
Publication Date
2024-11-08

AI Technical Summary

Technical Problem

The prior art has a small fatigue crack propagation rate and is difficult to distinguish between fatigue bands. The inverse pulsing depends on multiple assumptions and engineering experience, and the calculation workload is large and the efficiency is low, which affects the accuracy and aging of quantitative inverse pulsing of fatigue fractures. .

Method used

By conducting fatigue tests on metal material samples under different loading conditions, fracture morphology images, crack size and stress intensity factor range were obtained, training was used using the VGG-16 network model, hyperparameters were optimized, and the first and second prediction models were established. Used to predict the crack size and stress strength factor range of the material to be tested.

Benefits of technology

The accuracy and working efficiency of quantitative prediction of fatigue fractures are improved, and the relationship between fatigue fracture morphology and crack size and stress strength factor range can be more accurately described, reducing the calculation workload and improving the processing efficiency.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118918998A_ABST
    Figure CN118918998A_ABST
Patent Text Reader

Abstract

The invention discloses a metal material fatigue crack size and stress intensity factor range prediction method, and relates to the technical field of fatigue failure analysis, and the method comprises the following steps: carrying out fatigue tests on a metal material sample under different loading conditions to form different fatigue fractures; fracture morphology images at different positions on the fatigue fracture and corresponding crack sizes and stress intensity factor ranges are obtained; the VGG-16 network models are trained respectively; and respectively inputting the fracture morphology image of the to-be-predicted metal material into the first prediction model and the second prediction model to obtain the corresponding crack size and stress intensity factor range. According to the method provided by the invention, all detail information in the fatigue fracture morphology image is utilized to the greatest extent, and the relationship between the fatigue fracture morphology image and the corresponding crack size and stress intensity factor range can be described more accurately, so that the quantitative prediction accuracy of the fatigue fracture is improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention relates to the technical field of fatigue failure analysis, and in particular to a method for predicting the size of fatigue cracks and the range of stress intensity factors of metal materials. Background Art

[0002] Fatigue refers to the phenomenon that the performance of a material changes under repeated stress or strain. According to statistics, about 80% of failure and damage accidents of metal components are caused by fatigue fracture. Fatigue fracture is the matching surface morphology produced after fatigue damage of the material. It records the entire process of fatigue crack initiation and expansion before fatigue fracture of the material under external strain. Quantitative inversion of fatigue fracture is an important task in fatigue fracture failure analysis of metal materials. It not only helps to analyze the cause of fatigue fracture, obtain quantitative parameters such as the range of stress intensity factors and fatigue crack growth rate in the actual service of the material, but also provides important data and reference basis for in-depth analysis of the failure process. At the same time, by predicting the crack length, the fatigue life evaluation of parts can be realized. According to the crack length at different positions, the equivalent initial mass can be inferred, that is, the initial cracks or defects of the structural parts before service, and the material manufacturing and processing quality can be evaluated. Therefore, quantitative inversion of fatigue fracture fracture has important practical significance and engineering value.

[0003] At present, the quantitative inference of fatigue fracture is mainly based on the study of fatigue bands in fracture morphology. Through microscopic observation and calculation of the spacing between fatigue bands, the corresponding empirical formulas such as the Paris formula are used to infer quantitative parameters such as fatigue crack growth life and fatigue stress level. Although this method can obtain good inference results through calculation, it still has many limitations. For example, the spacing between fatigue bands is difficult to distinguish when the fatigue crack growth rate is small, the inference relies on multiple assumptions and engineering experience, and the inference calculation workload is large and the efficiency is low. These limitations seriously affect the accuracy and timeliness of quantitative inference of fatigue fracture. Summary of the invention

[0004] The present invention provides a method and system for predicting the range of fatigue crack size and stress intensity factor of metal materials, which solves the problems that the fatigue strip spacing is difficult to distinguish when the fatigue crack growth rate is small, the reverse calculation relies on multiple assumptions and engineering experience, the reverse calculation workload is large, and the efficiency is low.

[0005] The present invention provides a method for predicting the size and stress intensity factor range of fatigue cracks in metal materials, comprising the following steps:

[0006] Conduct fatigue tests on metal material samples under different loading conditions to form different fatigue fractures;

[0007] Obtain the fracture morphology images at different positions on the fatigue fracture and the corresponding crack size and stress intensity factor range;

[0008] The VGG-16 network model is trained by using multiple fracture morphology images and corresponding crack size and stress intensity factor ranges to obtain the first prediction model and the second prediction model;

[0009] Inputting the fracture morphology image of the metal material to be predicted into the first prediction model and the second prediction model respectively to obtain the corresponding crack size and stress intensity factor range;

[0010] The VGG-16 network model is trained by using multiple fracture morphology images and corresponding crack sizes and stress intensity factor ranges, including:

[0011] Inputting a plurality of fracture morphology images into two VGG-16 network models to obtain a plurality of first output results and a second output result;

[0012] Iterating a VGG-16 network model through crack sizes corresponding to a plurality of first output results and a plurality of fracture morphology images, and iterating another VGG-16 network model through stress intensity factor ranges corresponding to a plurality of second output results and a plurality of fracture morphology images;

[0013] In the iterative process, the ten-fold cross-validation algorithm is used to optimize the hyperparameters of the two VGG-16 network models to obtain the optimized first determination coefficient and second determination coefficient;

[0014] The hyperparameters corresponding to the maximum first determination coefficient and the maximum second determination coefficient are respectively selected as the model parameters of the first prediction model and the second prediction model.

[0015] Preferably, the VGG-16 network model includes:

[0016] The feature extraction module includes 13 convolutional layers and 5 maximum pooling layers, which are used to receive and extract the features of the fracture morphology image. Each convolutional layer uses a 3×3 convolution kernel, and each convolutional layer is followed by a ReLU activation function. The maximum pooling layer has a 2×2 pooling window.

[0017] The information integration module includes three fully connected layers, which is used to integrate the features extracted by the feature extraction module to obtain the integration results;

[0018] Output module, whose node number is 1, is used to output crack size or stress intensity factor range.

[0019] Preferably, before training the VGG-16 network model respectively with a plurality of fracture morphology images and corresponding crack size and stress intensity factor ranges, the fracture morphology images need to be cropped, randomly flipped and color gamut enhanced.

[0020] Preferably, it also includes:

[0021] A plurality of fracture morphology images and the corresponding crack size and stress intensity factor ranges are constructed as a data set;

[0022] The dataset is randomly divided into 80% training set and 20% test set;

[0023] The VGG-16 network model is trained using the training set to obtain a first prediction model and a second prediction model;

[0024] The test set is used to verify the accuracy of the first prediction model and the second prediction model.

[0025] Preferably, the fracture morphology images at different positions on the fatigue fracture of the metal material are obtained by a scanning electron microscope.

[0026] A method system for predicting fatigue crack size and stress intensity factor range of metal materials, comprising:

[0027] The test module is used to perform fatigue tests on metal material samples under different loading conditions to form different fatigue fractures;

[0028] An acquisition module is used to acquire fracture morphology images at different positions on the fatigue fracture and the corresponding crack size and stress intensity factor range;

[0029] A training module, used to train the VGG-16 network model through multiple fracture morphology images and corresponding crack size and stress intensity factor ranges to obtain a first prediction model and a second prediction model;

[0030] A prediction module, used to input the fracture morphology image of the metal material to be predicted into the first prediction model and the second prediction model respectively, to obtain the corresponding crack size and stress intensity factor range;

[0031] The training module includes:

[0032] An input module, used for inputting a plurality of fracture morphology images into two VGG-16 network models to obtain a plurality of first output results and a second output result;

[0033] An iteration module, for iterating a VGG-16 network model through a plurality of first output results and crack sizes corresponding to a plurality of fracture morphology images, and iterating another VGG-16 network model through a plurality of second output results and stress intensity factor ranges corresponding to a plurality of fracture morphology images;

[0034] An optimization module is used to optimize the hyperparameters of two VGG-16 network models using a ten-fold cross-validation algorithm during the iteration process to obtain the optimized first determination coefficient and the second determination coefficient;

[0035] The selection module is used to respectively select the hyperparameters corresponding to the maximum first determination coefficient and the maximum second determination coefficient as the model parameters of the first prediction model and the second prediction model.

[0036] A computer device comprises a memory, a processor and a computer program stored in the memory and executable on the processor. When the processor executes the program, the method for predicting the range of fatigue crack size and stress intensity factor of metal material is realized.

[0037] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the above-mentioned method for predicting the range of fatigue crack size and stress intensity factor of a metal material.

[0038] Compared with the prior art, the present invention has the following beneficial effects:

[0039] The present invention firstly performs fatigue tests on metal material samples under different loading conditions to form different fatigue fractures, then obtains fracture morphology images at different positions on the fatigue fracture and the corresponding crack size and stress intensity factor range, trains the convolutional neural network model through the obtained fracture morphology images and the corresponding crack size and stress intensity factor range, optimizes the hyperparameters of the two VGG-16 network models by using a ten-fold cross validation algorithm during the training process, selects the hyperparameters corresponding to the maximum first determination coefficient and the second determination coefficient as the model parameters of the first prediction model and the second prediction model, obtains the first prediction model and the second prediction model, inputs the fatigue fracture morphology image of the metal material to be predicted into the first prediction model and the second prediction model, and predicts the crack size and stress intensity factor range. Compared with the traditional method of inferring the crack size and stress intensity factor range by measuring the fatigue fracture band spacing, the method proposed by the present invention makes the most of all the detailed information in the fatigue fracture morphology image, can more accurately describe the relationship between the fatigue fracture morphology image and the corresponding crack size and stress intensity factor range, thereby improving the accuracy of quantitative prediction of fatigue fracture. In addition, the method proposed in the present invention can process a large number of fatigue fracture morphology images in a short time, greatly improving work efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0041] Figure 1 A flowchart of a method for predicting the range of fatigue crack size and stress intensity factor of a metal material according to the present invention;

[0042] Figure 2 It is a fracture morphology image of an embodiment of the present invention;

[0043] Figure 3 This is a cropped fracture morphology image of an embodiment of the present invention;

[0044] Figure 4 A schematic diagram of crack size prediction results of a training set according to an embodiment of the present invention;

[0045] Figure 5 This is a schematic diagram of the prediction results of the range of stress intensity factors of the training set according to an embodiment of the present invention;

[0046] Figure 6 This is a schematic diagram of crack size prediction results of a test set according to an embodiment of the present invention;

[0047] Figure 7 Schematic diagram of the prediction results of the stress intensity factor range of the test set according to an embodiment of the present invention. DETAILED DESCRIPTION

[0048] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0049] The present invention discloses a method for predicting the fatigue crack size and stress intensity factor range of a metal material. The present invention first obtains the fracture morphology images at different positions on the fatigue fracture of the metal material under various loading conditions by means of a scanning electron microscope. Next, a fatigue fracture quantitative prediction data set is constructed, and each set of data includes a fatigue fracture morphology image and a corresponding crack size and stress intensity factor range, and the data set is randomly divided into a training set and a test set. Thirdly, the training set data is used to construct a fatigue fracture quantitative prediction model using a convolutional neural network model, and the model parameters are optimized, and the model accuracy is evaluated using the test set data. Finally, the newly obtained fatigue fracture morphology image to be predicted is input into the fatigue fracture quantitative prediction model to predict its crack size and stress intensity factor range. Specifically, refer to Figure 1 , including the following steps:

[0050] The first step: obtain metal material samples, conduct fatigue tests on the metal material samples under different loading conditions, and form different fatigue fractures.

[0051] Step 2: Obtain the fracture morphology images at different positions on the fatigue fracture and the corresponding crack size and stress intensity factor range.

[0052] The image name is marked in the format of "material-maximum tensile force-load ratio-crack length-magnification-(serial number)". In this embodiment, the crack size is the crack length. The stress intensity factor range can be calculated by national standards or theoretical calculation formulas.

[0053] Step 3: Train the VGG-16 network model using multiple fracture morphology images and the corresponding crack size and stress intensity factor ranges to obtain the first prediction model and the second prediction model.

[0054] The network structure of VGG-16 is divided into three parts: feature extraction, information integration, and output. The feature extraction part consists of 13 convolutional layers and 5 maximum pooling layers, which are used to gradually extract the features of the input image. Each convolutional layer uses a 3×3 convolution kernel and a smaller stride and padding, which can gradually increase the depth of the network while maintaining the image dimension. After each convolutional layer, there is a ReLU activation function, which helps the network learn nonlinear features. The formula of the ReLU function is as follows:

[0055] f(x)=max(x,0)

[0056] The maximum pooling layer is located between multiple convolutional layers and is used to reduce the dimension of the feature map and reduce the amount of calculation. The maximum pooling layer used by VGG-16 is a 2×2 pooling window with a stride of 2. This design can halve the size of the feature map while retaining the most important features. After the feature extraction part is the information integration part, which includes three fully connected layers, which are used to integrate the features extracted by the feature extraction part to form the final classification or regression results. The output part corresponds to the last fully connected layer (output layer) of VGG-16. The number of nodes in the output layer is 1, and the output is a fixed value, that is, the range of crack size or stress intensity factor that needs to be predicted.

[0057] The data set is constructed with the fracture morphology images obtained in the first step. The input features of each set of data include the fracture morphology image, and the output features include the crack size and stress intensity factor range. The data set is randomly divided into 80% training set and 20% test set.

[0058] Each fracture morphology image in the dataset is evenly cropped into 6 images to increase the number of samples and enable the model to learn image features more fully.

[0059] The cropped images are preprocessed before model training, including random flipping and color gamut enhancement, to improve the generalization performance of the model.

[0060] The training set data is used to train the deep learning model for predicting the quantitative parameters of fatigue fracture, namely the first prediction model and the second prediction model, through the convolutional neural network VGG-16 algorithm.

[0061] The ten-fold cross-validation algorithm was used to optimize the hyperparameters of the convolutional neural network model, and the determination coefficient of the optimized convolutional neural network model was calculated to evaluate the model training effect.

[0062] Select the coefficient of determination R 2 The model hyperparameters corresponding to the maximum value of Train are used as the model parameters of the first prediction model and the second prediction model.

[0063] The coefficient of determination R used to describe the model training effect 2 Train is as follows:

[0064]

[0065] Where n1 is the number of fracture morphology images in the training set, y i is the actual value of the crack size or stress intensity factor range corresponding to the image, is the predicted value of the crack size or stress intensity factor range corresponding to the image, is the average of the actual values.

[0066] The accuracy of the established first prediction model and the second prediction model is verified based on the input data in the test set.

[0067] Using the first prediction model and the second prediction model, the predicted values ​​of crack size and stress intensity factor range are obtained by using the fracture morphology image to be predicted in the test set, and compared with the corresponding true values. The root mean square error R used to describe the model test effect is calculated. 2 Test.

[0068] The root mean square error R that describes the model testing effect 2 Test is:

[0069]

[0070] Where n2 is the number of fracture morphology images in the test set, y i is the actual value of the crack size or stress intensity factor range corresponding to the image, is the predicted value of the crack size or stress intensity factor range corresponding to the image, is the average of the actual values.

[0071] Step 4: Input the fracture morphology image of the metal material to be predicted into the first prediction model and the second prediction model respectively to obtain the corresponding crack size and stress intensity factor range.

[0072] Based on the same concept, the present invention also provides a metal material fatigue fracture quantitative prediction system based on deep learning, including a test module, an acquisition module, a training module and a prediction module.

[0073] The test module is used to perform fatigue tests on metal material samples under different loading conditions to form different fatigue fractures.

[0074] The acquisition module is used to obtain the fracture morphology images at different positions on the fatigue fracture and the corresponding crack size and stress intensity factor range.

[0075] The training module is used to train the VGG-16 network model through multiple fracture morphology images and corresponding crack size and stress intensity factor ranges to obtain a first prediction model and a second prediction model.

[0076] The prediction module is used to input the fracture morphology image of the metal material to be predicted into the first prediction model and the second prediction model respectively to obtain the corresponding crack size and stress intensity factor range.

[0077] The training module includes an input module, an iteration module, an optimization module and a selection module.

[0078] The input module is used to input multiple fracture morphology images into two VGG-16 network models to obtain multiple first output results and second output results.

[0079] The iteration module is used to iterate a VGG-16 network model through crack sizes corresponding to multiple first output results and multiple fracture morphology images, and to iterate another VGG-16 network model through stress intensity factor ranges corresponding to multiple second output results and multiple fracture morphology images.

[0080] The optimization module is used to optimize the hyperparameters of the two VGG-16 network models using a ten-fold cross-validation algorithm during the iteration process to obtain the optimized first determination coefficient and second determination coefficient.

[0081] The selection module is used to select the hyperparameters corresponding to the maximum first determination coefficient and the maximum second determination coefficient as the model parameters of the first prediction model and the second prediction model respectively.

[0082] The present invention also provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, the above-mentioned method for predicting the range of fatigue crack size and stress intensity factor of metal materials is implemented.

[0083] The present invention also provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, the above-mentioned method for predicting the range of fatigue crack size and stress intensity factor of metal materials is implemented.

[0084] Example 1

[0085] Refer again Figure 1 , this embodiment includes the following steps:

[0086] Step 1: Obtain fracture morphology images at different positions on the fatigue fracture of metal materials under various loading conditions and mark them with names.

[0087] The metal material used in this embodiment is 2.25Cr1MoV high-strength steel, and a total of 80 images of morphology of fatigue fractures at different positions (crack sizes) under different loading conditions (as shown in Table 1) were obtained using a scanning electron microscope. The image names are marked in the format of "material-maximum tensile force-load ratio-crack length-magnification-serial number", "225Cr-20-0.1-5mm-5k-2". Figure 2 This is one of the fracture morphology images of an embodiment of the present invention.

[0088] Table 1 Fatigue loading conditions of 2.25Cr1MoV high strength steel

[0089]

[0090] Step 2: The convolutional neural network model is trained by acquiring the fracture morphology image and the corresponding crack size and stress intensity factor range to obtain the first prediction model and the second prediction model.

[0091] The fatigue fracture quantitative prediction data set is constructed with the fracture morphology images obtained in step 1. The input features of each set of data include the fracture morphology image, and the output features include the crack size and stress intensity factor range. The data set is randomly divided into 80% training set and 20% test set. Therefore, the training set in this embodiment contains 64 images and the test set contains 16 images.

[0092] Each fracture morphology image in the dataset is evenly cropped into 6 images to increase the number of samples and enable the model to learn image features more fully. Figure 3 This is a cropped fracture morphology image of an embodiment of the present invention.

[0093] The cropped images are preprocessed before model training, including random flipping and color gamut enhancement to improve the generalization performance of the model;

[0094] The training set data was used to train a deep learning model for predicting quantitative parameters of fatigue fracture through the convolutional neural network VGG-16 algorithm.

[0095] The 10-fold cross validation algorithm was used to optimize the hyperparameters of the VGG-16 model, and the determination coefficient of the optimized convolutional neural network model was calculated to evaluate the model training effect; the determination coefficient R was selected 2 The model hyperparameters corresponding to the maximum value of Train are used as the model parameters of the VGG-16 model. The VGG-16 model parameters are shown in Table 2.

[0096] Table 2 VGG-16 model hyperparameter results

[0097]

[0098] Figure 4 and Figure 5 The prediction results of fatigue fracture crack size and stress intensity factor range of 2.25Cr1MoV high-strength steel based on VGG-16 model using training set data are given respectively. For the prediction of crack size, the determination coefficient R of 2.25Cr1MoV high-strength steel samples in the training set is 2 Train is 0.82, showing a good fitting effect, indicating that the model has fully learned the characteristics of the image; for the prediction problem of the stress intensity factor range, the R of the 2.25Cr1MoV high-strength steel training set is 2 The Train value is 0.82, which is a good result.

[0099] The accuracy of the first and second prediction models established was verified based on the input data in the test set. The first and second prediction models were used to obtain the predicted values ​​of the crack size and stress intensity factor range using the fracture morphology images in the test set, and compared with the corresponding true values. Figure 6 and Figure 7 The prediction results of fatigue fracture crack size and stress intensity factor range of 2.25Cr1MoV high-strength steel based on VGG-16 model using test set data are given respectively. For the prediction of crack size, the determination coefficient R of 2.25Cr1MoV high-strength steel samples on the test set is 2 Test is 0.52, which shows a high prediction accuracy. At the same time, the prediction results are generally large and conservative, which helps to improve the safety in engineering applications. 2 The test is 0.53, which achieves a good fitting effect and can complete the prediction task well.

[0100] Step 3: Input the newly obtained fracture morphology image to be predicted into the first prediction model and the second prediction model to predict the crack size and stress intensity factor range. Although the preferred embodiments of the present invention have been described, those skilled in the art may make additional changes and modifications to these embodiments once they are aware of the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present invention.

[0101] Obviously, those skilled in the art may make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. A method for predicting fatigue crack size and stress intensity factor range of metal materials, characterized in that: The following steps are involved: Conduct fatigue tests on metal material samples under different loading conditions to form different fatigue fractures; Obtain the fracture morphology images at different positions on the fatigue fracture and the corresponding crack size and stress intensity factor range; The VGG-16 network model is trained by using multiple fracture morphology images and corresponding crack size and stress intensity factor ranges to obtain the first prediction model and the second prediction model; Inputting the fracture morphology image of the metal material to be predicted into the first prediction model and the second prediction model respectively to obtain the corresponding crack size and stress intensity factor range; The VGG-16 network model is trained by using multiple fracture morphology images and corresponding crack sizes and stress intensity factor ranges, including: Inputting a plurality of fracture morphology images into two VGG-16 network models to obtain a plurality of first output results and a second output result; Iterating a VGG-16 network model through crack sizes corresponding to a plurality of first output results and a plurality of fracture morphology images, and iterating another VGG-16 network model through stress intensity factor ranges corresponding to a plurality of second output results and a plurality of fracture morphology images; In the iterative process, the ten-fold cross-validation algorithm is used to optimize the hyperparameters of the two VGG-16 network models to obtain the optimized first determination coefficient and second determination coefficient; The hyperparameters corresponding to the maximum first determination coefficient and the maximum second determination coefficient are respectively selected as the model parameters of the first prediction model and the second prediction model.

2. A method for predicting fatigue crack size and stress intensity factor range of metal materials according to claim 1, characterized in that: The VGG-16 network model includes: The feature extraction module includes 13 convolutional layers and 5 maximum pooling layers, which are used to receive and extract the features of the fracture morphology image. Each convolutional layer uses a 3×3 convolution kernel, and each convolutional layer is followed by a ReLU activation function. The maximum pooling layer has a 2×2 pooling window. The information integration module includes three fully connected layers, which is used to integrate the features extracted by the feature extraction module to obtain the integration results; Output module, whose node number is 1, is used to output crack size or stress intensity factor range.

3. A method for predicting fatigue crack size and stress intensity factor range of metal materials according to claim 1, characterized in that: Before training the VGG-16 network model using multiple fracture morphology images and the corresponding crack size and stress intensity factor ranges, the fracture morphology images need to be cropped, randomly flipped, and color gamut enhanced.

4. A method for predicting fatigue crack size and stress intensity factor range of metal materials according to claim 1, characterized in that: Also includes: A plurality of fracture morphology images and the corresponding crack size and stress intensity factor ranges are constructed as a data set; The dataset is randomly divided into 80% training set and 20% test set; The VGG-16 network model is trained using the training set to obtain a first prediction model and a second prediction model; The test set is used to verify the accuracy of the first prediction model and the second prediction model.

5. A method for predicting fatigue crack size and stress intensity factor range of metal materials according to claim 1, characterized in that: The fracture morphology images at different positions on the fatigue fracture of metal materials are obtained by scanning electron microscopy.

6. A method and device for predicting the range of fatigue crack size and stress intensity factor of metal materials, characterized in that: include: The test module is used to perform fatigue tests on metal material samples under different loading conditions to form different fatigue fractures; An acquisition module is used to acquire fracture morphology images at different positions on the fatigue fracture and the corresponding crack size and stress intensity factor range; A training module, used to train the VGG-16 network model through multiple fracture morphology images and corresponding crack size and stress intensity factor ranges to obtain a first prediction model and a second prediction model; A prediction module, used for inputting the fracture morphology image of the metal material to be predicted into the first prediction model and the second prediction model respectively, to obtain the corresponding crack size and stress intensity factor range; The training module includes: An input module, used for inputting a plurality of fracture morphology images into two VGG-16 network models to obtain a plurality of first output results and a second output result; An iteration module, for iterating a VGG-16 network model through a plurality of first output results and crack sizes corresponding to a plurality of fracture morphology images, and iterating another VGG-16 network model through a plurality of second output results and stress intensity factor ranges corresponding to a plurality of fracture morphology images; An optimization module is used to optimize the hyperparameters of two VGG-16 network models using a ten-fold cross-validation algorithm during the iteration process to obtain the optimized first determination coefficient and the second determination coefficient; The selection module is used to respectively select the hyperparameters corresponding to the maximum first determination coefficient and the maximum second determination coefficient as the model parameters of the first prediction model and the second prediction model.

7. A computer device, characterized in that: The method comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the method for predicting the range of fatigue crack size and stress intensity factor of metal materials as described in any one of claims 1 to 5 is implemented.

8. A computer-readable storage medium, characterized in that: The storage medium stores a computer program, and when the computer program is executed by a processor, the method for predicting the range of fatigue crack size and stress intensity factor of a metal material as described in any one of claims 1 to 5 is implemented.

Citation Information

Patent Citations

  • Fatigue crack growth rate prediction method based on artificial neuron network

    CN105488328A

  • Method for estimating crack initiation dimension values

    CN110059432A

  • Road surface crack image detection method

    CN110120041A

  • Model training method, metal fracture analysis method based on deep learning and application

    CN111209964A

  • Fatigue fracture machine vision image intelligent analysis method, device and system

    CN111681217A

Cited By

  • Online prediction method and equipment for hot cracks of brake disc based on microstructure transfer learning and medium

    CN121258979A

  • Intelligent identification method and system for morphological characteristics of fracture of metal material

    CN121564714A

  • Metal material fracture morphology feature intelligent identification method and system thereof

    CN121564714B