A method for identifying damage patterns of carbon fiber reinforced composites using interpretable convolutional neural networks

Abstract

This invention discloses a method for identifying damage patterns in carbon fiber reinforced composite materials using an interpretable convolutional neural network. The method comprises the following steps: establishing a database; preprocessing data; constructing a convolutional neural network model; interpreting the convolutional neural network model; and generalizing the convolutional neural network model. After model training, the model is generalized and applied to identify multiple damage pattern images and composite impact damage patterns in actual engineering projects to verify the robustness of the model. The technical solution provided by this invention can accurately and efficiently identify damage patterns in carbon fiber reinforced composite materials, achieving precise quantification and efficient decoupling of composite material damage patterns under service conditions.

Description

Technical Field

[0001] The present invention relates to the technical field of composite material damage detection, and in particular to a method for identifying damage patterns of carbon fiber reinforced composite materials using an interpretable convolutional neural network. Background Art

[0002] Carbon fiber reinforced composite materials refer to composite materials that use carbon fiber or carbon fiber fabric as reinforcement and are formed with a matrix such as resin, ceramic, metal, cement, carbon or rubber. Due to its advantages such as high specific strength, high specific stiffness, corrosion resistance and strong designability, it is widely used in aerospace, civil construction, chemical equipment, medical equipment and other fields. However, since carbon fiber reinforced composite materials are usually used in harsh environments and complex load conditions, they are inevitably subjected to impact loads, causing various types of damage. [1] .

[0003] The damage sensitivity and severity of composite materials can be affected by many factors, including the material factors, manufacturing factors, load and environmental factors, design factors, etc. [2, 3] Carbon fiber reinforced composite materials have complex failure mechanisms. Common damage to carbon fiber reinforced composite materials includes matrix cracking, fiber-matrix interface peeling, delamination, fiber extraction and fiber breakage, which can easily lead to material performance degradation and structural damage, affecting its service status throughout its life cycle. [4-6] .

[0004] Currently, the traditional damage detection methods for composite materials include ultrasonic nondestructive testing [7] , Acoustic Emission Testing [8] , optical nondestructive testing [9] , infrared thermal imaging detection

[10] and thermal imaging technology

[11] However, traditional non-destructive testing technologies have their own scope of application and certain limitations.

[12] For example, ultrasonic testing technology is not suitable for complex service conditions.

[13] Acoustic emission technology requires good coupling, expensive equipment, and certain technical requirements for testers.

[14] Optical nondestructive testing technology requires materials with good heat absorption rate and high depth requirements. The deeper the depth, the lower the detection sensitivity. It is suitable for relatively thin laminates, and the edge imaging is easily blurred during detection.

[15] In order to overcome the shortcomings of existing technologies, it is necessary to study a composite material damage detection method with wide applicability, high reliability and strong robustness, improve the efficiency and accuracy of composite material structure damage detection, and improve the applicability of damage detection. Summary of the Invention

[0005] The present invention provides a method and device for predicting dynamic stress at structural defects based on a BP neural network, aiming to effectively solve the technical problems faced in the prior art, such as the time-consuming calculation of multiple parameters, poor applicability when dealing with different structures, materials and defect forms, resulting in high time and labor costs.

[0006] To address the shortcomings of existing technologies, the present invention aims to provide a method for identifying damage patterns in carbon fiber reinforced composite materials using an interpretable convolutional neural network. This method addresses the high cost, low accuracy, and inability of current composite material damage detection methods to characterize the extent and type of internal damage under low-velocity impact loads. Furthermore, addressing the "black box" problem of machine learning models, the present invention provides a method for interpreting the internal workings of convolutional neural network models, effectively identifying and preventing potential biases.

[0007] According to a first aspect of the present invention, the present invention provides a method for identifying damage modes of carbon fiber reinforced composite materials using an interpretable convolutional neural network, comprising the following steps:

[0008] Establish a database to collect and organize CT images of carbon fiber composite laminates damaged under low-speed impact loads and establish a CT damage pattern image dataset;

[0009] Data preprocessing: resizing the image size to a preset size and adjusting the mean and standard deviation of the computed tomography damage pattern image dataset for standardization;

[0010] Constructing a convolutional neural network model, wherein the convolutional neural network model includes a convolutional layer and a fully connected layer;

[0011] Training and testing convolutional neural network models, using computed tomography damage pattern image datasets to train and test the established convolutional neural network models, and performing classification, recognition, and quantitative analysis on carbon fiber composite material impact damage pattern image datasets;

[0012] Interpretation of convolutional neural network models: introducing interpretable tools such as class activation maps and Shapley additive explanatory values ​​to transparently explain convolutional neural network models, reveal the underlying variables and mechanisms that influence model predictions, visualize the contribution of each type of damage pattern to the final prediction results, and perform rotational validation on a computed tomography damage image dataset.

[0013] Generalization application of the convolutional neural network model. After the model training is completed, the model is generalized and applied to identify multiple damage pattern images and composite material impact damage patterns in actual engineering to verify the robustness of the model.

[0014] In addition to or as an alternative to one or more of the features disclosed above, the method further comprises:

[0015] Composite material preparation steps:

[0016] - Using carbon fiber or carbon fiber fabric as reinforcement and resin as matrix;

[0017] -Preparing carbon fiber composite laminates by a lamination process;

[0018] Sample preparation steps:

[0019] -Select carbon fiber composite laminates as sample plates;

[0020] -Manufacture multiple sets of specimens according to predetermined specifications, where the shapes and widths of the specimens in different sets are consistent, but the thicknesses vary;

[0021] - Prepare two identical parallel samples for each set of specimens;

[0022] Impact test steps:

[0023] -Choice of two shock sources with different energy levels;

[0024] - Two parallel samples of each group of specimens are placed under impact sources of different energies for impact tests;

[0025] Data processing steps:

[0026] - Use computed tomography to examine the specimens after impact testing;

[0027] - Obtain computed tomography damage image data of the damaged area under different impact energy loads.

[0028] In addition to or as an alternative to one or more of the features disclosed above, the damage mode of the carbon fiber composite laminate under load impact of different energy levels includes at least one of matrix cracking, delamination or fiber breakage.

[0029] In addition to one or more of the features disclosed above, or as an alternative, the computed tomography damage image dataset is divided into a single damage pattern dataset and a multiple damage pattern dataset, wherein the image data in the single damage pattern dataset contains only one damage pattern, and the image data in the multiple damage pattern dataset contains two or more damage patterns.

[0030] In addition to or as an alternative to one or more of the features disclosed above, the step of adjusting the mean and standard deviation of the computed tomography lesion pattern image dataset for normalization further comprises:

[0031] Each computed tomography damage pattern image data was transformed into a normal distribution with a mean of 1 and a variance of 0.

[0032] In addition to or as an alternative to one or more of the features disclosed above, the convolutional layer is used to perform damage edge detection and feature extraction on the computed tomography damage image, and the fully connected layer is used to highly abstract and integrate the features extracted by the convolutional layer to output the final laminate damage pattern recognition result;

[0033] Among them, there are 5 convolutional layers and 3 fully connected layers; the network structure of the convolutional neural network model also includes 3 pooling layers and 1 normalized exponential function layer.

[0034] In addition to or instead of one or more of the features disclosed above, the number of channels of the first to fifth convolutional layers of the convolutional neural network model are 16, 32, 64, 96, and 128, respectively.

[0035] In addition to or instead of one or more of the features disclosed above, the learning rate of the convolutional neural network model adopts an adaptive adjustment strategy with an initial learning rate of 1e-4 and a decay rate of 0.1. The learning rate changes every 10 epochs, so that the model can adapt more quickly to the distribution characteristics of various damage patterns in the target domain.

[0036] In addition to one or more of the features disclosed above, or as an alternative, the steps of training and testing the established convolutional neural network model using the computed tomography damage pattern image dataset include: randomly selecting 80 samples from each of the three types of damage patterns in the standardized computed tomography single damage pattern image dataset to form a test set, and randomly dividing the remaining computed tomography single damage pattern images into a training set and a validation set in a ratio of 4:1; using the training set and the validation set to train the convolutional neural network model, and adjusting the hyperparameter settings of the network model during the training process; after the model training is completed, using the test set to test the damage pattern recognition performance of the convolutional neural network model for the computed tomography images.

[0037] In addition to or as an alternative to one or more of the features disclosed above, for the trained convolutional neural network model, a class activation map is introduced to reweight the weights of the last convolutional layer of the convolutional neural network model, visualize the importance of each input feature, visualize the decision process of the proposed convolutional neural network model, and explain the reasons behind the model's decision.

[0038] In addition to or as an alternative to one or more features disclosed above, the class activation map includes at least one of Grad-CAM, Grad-CAM++, Score-CAM, XGrad-CAM, Ablation-CAM, EigenGrad-CAM, and LayerCAM.

[0039] In addition to one or more of the features disclosed above, or as an alternative, Shapley additive explanatory value analysis is introduced for the trained convolutional neural network model to calculate the contribution of different damage pattern features to the prediction results of the convolutional neural network model, provide an explanatory score for each damage pattern feature, and reveal the impact of different damage patterns on the model output.

[0040] In addition to one or more of the features disclosed above, or as an alternative, for the trained convolutional neural network model, the three damage pattern images of matrix cracking, delamination, and fiber breakage in the test set are randomly assigned to be rotated 90°, 180°, and 270° clockwise in sequence. The class activation map and Shapley additive explanatory value analysis visualization model are used to identify and predict the three damage pattern features after rotation.

[0041] In addition to or instead of one or more of the features disclosed above, the hyperparameters and weight values ​​of each layer of the optimal convolutional neural network model are stored, and the network model is called to identify a computed tomography multiple damage pattern data set; the model is applied to actual engineering to identify different damage modes of carbon fiber reinforced composite laminates under low-speed impact.

[0042] In addition to or as an alternative to one or more of the features disclosed above, multiple damage pattern images are efficiently decoupled and quantitatively analyzed, and interpretable tools such as class activation maps and Shapley additive explanatory values ​​are introduced for transparent interpretation, revealing the potential variables and mechanisms that affect the model prediction results, and visualizing the contribution of each type of damage pattern to the final prediction results; confidence values ​​are introduced to quantitatively analyze the prediction results.

[0043] In addition to or as an alternative to one or more features disclosed above, the verifying the initial dynamic stress prediction model using the test set includes the following steps:

[0044] Keeping the parameters of the initial dynamic stress prediction model unchanged, each test sample in the test set is applied to the initial dynamic stress prediction model for verification;

[0045] If the sum of the test errors of each sample in the test set reaches the preset error tolerance, it indicates that the initial dynamic stress prediction model has good generalization ability and the test passes; if the sum of the test errors is too large, underfitting or overfitting occurs, the test fails, and the training set is reused to perform initial training on the BP neural network prediction model.

[0046] One of the above technical solutions has the following advantages or beneficial effects: the convolutional neural network model is trained and tested through a single damage data set of computed tomography (CT), and the hyperparameters and weight values ​​of the convolutional neural network model after training are stored. It can be applied to multiple damage pattern data sets and other damage pattern recognition tasks of carbon fiber reinforced composite materials under impact loads, and has strong robustness.

[0047] Another technical solution among the above technical solutions has the following advantages or beneficial effects: due to the introduction of Class Activation Mapping (CAM) technology and Shapley Additive Interpretation (SHAP) value interpretable tools to transparently interpret the convolutional neural network, CAM performs local interpretation of the convolutional neural network, which well illustrates the reliability of the convolutional neural network model in black-box prediction of regions of interest in different damage patterns. SHAP calculates the marginal contribution of the characteristics of different damage patterns to the model prediction results, reveals the potential variables and mechanisms that affect the model prediction results, and truly establishes trust between people and models.

[0048] Another technical solution among the above technical solutions has the following advantages or beneficial effects: the constructed convolutional neural network model comprehensively considers the importance of convolutional layers and channels, contains 5 convolutional layers, and has a simple and lightweight structure; the network model adopts a larger number of channels in the early stage, which can process more input data information at the same time, extract more features of different damage modes of composite materials, and improve the accuracy in damage pattern recognition tasks. The network model adopts a smaller number of channels in the later stage and introduces the dropout method in the fully connected layer to effectively prevent the convolutional neural network model from overfitting. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] The technical solutions and other beneficial effects of the present invention will be made apparent by describing in detail the specific embodiments of the present invention in conjunction with the accompanying drawings.

[0050] Figure 1 is a schematic diagram of a computed tomography scan provided according to an embodiment of the present invention;

[0051] Figure 2 is a schematic diagram of a process provided according to an embodiment of the present invention;

[0052] Figure 3 2 is a schematic diagram of the structure of an interpretable convolutional neural network model provided according to an embodiment of the present invention;

[0053] Figure 4 2. It is a schematic diagram of the damage pattern recognition result of an interpretable convolutional neural network provided by an embodiment of the present invention;

[0054] Figure 5 1. It is a schematic diagram of a convolutional neural network model explained using the class activation map technology provided by an embodiment of the present invention;

[0055] Figure 6 1 is a schematic diagram of a convolutional neural network model for analyzing and explaining the Shapley additive interpretation (SHAP) value according to an embodiment of the present invention. DETAILED DESCRIPTION

[0056] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without making any creative efforts shall fall within the scope of protection of the present invention.

[0057] In the description of the present invention, it should be noted that, unless otherwise specified or limited, the term "and / or" herein is merely a description of an association relationship between associated objects, indicating that three possible relationships exist. For example, "A and / or B" can represent: A exists alone, A and B exist simultaneously, and B exists alone. Furthermore, the character " / " herein, unless otherwise specified, generally indicates that the associated objects are in an "or" relationship.

[0058] Reference Figure 1 In one embodiment of the present invention, a method for identifying damage patterns of carbon fiber reinforced composite materials using an interpretable convolutional neural network is disclosed, which may include the following steps:

[0059] T01, database establishment, collects and organizes CT images of carbon fiber composite laminates damaged under low-speed impact loads, and establishes a CT damage pattern image dataset;

[0060] T02, data preprocessing, resizing the image size to a preset size, adjusting the mean and standard deviation of the computed tomography damage pattern image dataset for standardization;

[0061] T03, constructing a convolutional neural network model, wherein the convolutional neural network model includes a convolutional layer and a fully connected layer;

[0062] T04, training and testing convolutional neural network models, using computed tomography damage pattern image datasets to train and test the established convolutional neural network models, and performing classification, recognition, and quantitative analysis on carbon fiber composite material impact damage pattern image datasets;

[0063] T05, Interpreting Convolutional Neural Network Models, introduces interpretable tools such as class activation maps and Shapley additive explanatory values ​​to transparently explain convolutional neural network models, revealing the underlying variables and mechanisms that influence model predictions. It also visualizes the contribution of each damage pattern to the final predictions and performs rotation validation on a computed tomography damage image dataset. Rotational transformations alter the location and orientation of cracks to varying degrees, further validating the model's robustness and improving its interpretability, enabling understanding of the model's decision-making process and key features.

[0064] T06, generalization application of convolutional neural network model. After the model training is completed, the model will be generalized and applied to identify multiple damage pattern images and composite material impact damage patterns in actual engineering to verify the robustness of the model.

[0065] It is understandable that the order of step T01 and step T03 can be changed. Step T01 can be executed first and then step T03, or step T03 can be executed first and then step T01, or step T01 and step T03 can be executed at the same time. The specific execution order can be set according to actual conditions and is not limited here.

[0066] It is also understandable that the order of step T02 and step T03 can be changed. Step T02 can be executed first and then step T03, or step T03 can be executed first and then step T02, or step T02 and step T03 can be executed at the same time. The specific execution order can be set according to actual conditions and is not limited here.

[0067] Computed tomography (CT) can provide non-destructive 3D images of an object's internal structure. CT uses the penetrating power of X-rays to obtain a series of 2D radiographs of an object viewed from multiple angles, and then computationally creates a series of cross-sectional images to reveal the object's internal structure. Non-destructive CT has been used to image delicate samples that are difficult to section (such as frozen ice cream) or samples that should not be damaged (such as cultural relics), and to assess the integrity of engineering components (such as turbine blades). CT can also be used for intermittent or continuous time-dependent studies to monitor the evolution of three-dimensional structures, such as the growth of malignant tumors, the metamorphosis of pupae, the flow of fluids in rocks, or the catastrophic thermal runaway failure of lithium batteries. CT was the first method capable of producing images in transverse planes (slices) through the human body, without distortion caused by the superposition of different anatomical structures. Since the first clinical CT scanner, significant technological advances in X-ray sources and new detectors have led to significant improvements in transmitted dose and image quality. CT technology has evolved to higher-resolution devices. X-ray micro-tomography (μ-CT) systems can achieve micron-level spatial resolution for centimeter-sized samples and submicron-level spatial resolution (nano-tomography) for samples 1-2 mm in size. Today, this technology is used in a wide range of research fields, from medicine to biology, from geology to archaeology and materials science.

[0068] Class activation maps (CAMs) are a visualization technique primarily used in convolutional neural networks for deep learning to identify and explain which regions in images or time series data are most critical to the model's decision-making process. They are particularly useful for understanding how a network makes classification decisions based on visual or time series features. Shapley additive explanations (SHAPs) are an interpretability method based on Shapley values ​​that aims to explain the outputs of machine learning models. SHAPs provide both global and local interpretability by calculating the contribution of each feature to the model's predictions.

[0069] The present invention trains and tests a convolutional neural network model using a single damage dataset from computed tomography (CT), and stores the hyperparameters and weight values ​​of each layer of the convolutional neural network model after training. This model can be applied to multiple damage pattern datasets and other damage pattern recognition tasks of carbon fiber reinforced composite materials under impact loads, and has strong robustness.

[0070] Furthermore, the introduction of Class Activation Mapping (CAM) technology and Shapley Additive Explanation (SHAP) value interpretable tools provide transparent interpretation of convolutional neural networks. CAM provides local interpretation of convolutional neural networks, which well illustrates the reliability of convolutional neural network models in black-box prediction of regions of interest with different damage patterns. SHAP calculates the marginal contribution of the features of different damage patterns to the model prediction results, revealing the potential variables and mechanisms that affect the model prediction results, and truly building trust between people and models.

[0071] In order to collect and organize the CT images of carbon fiber composite laminates damaged under low-speed impact loads, some pre-processing is usually required:

[0072] S01, composite material preparation steps:

[0073] - Using carbon fiber or carbon fiber fabric as reinforcement and resin as matrix;

[0074] -Preparing carbon fiber composite laminates by a lamination process;

[0075] S02, sample preparation steps:

[0076] -Select carbon fiber composite laminates as sample plates;

[0077] -Manufacture multiple sets of specimens according to predetermined specifications, where the shapes and widths of the specimens in different sets are consistent, but the thicknesses vary;

[0078] - Prepare two identical parallel samples for each set of specimens;

[0079] S03, impact test steps:

[0080] -Choice of two shock sources with different energy levels;

[0081] - Two parallel samples of each group of specimens are placed under impact sources of different energies for impact tests;

[0082] S04, data processing steps:

[0083] - Use computed tomography to examine the specimens after impact testing;

[0084] - Obtain computed tomography damage image data of the damaged area under different impact energy loads.

[0085] In one embodiment, four groups of carbon fiber composite laminate specimens of different thicknesses were produced according to a specification of 100 mm × 100 mm. The thicknesses of the four groups of carbon fiber composite laminate specimens were 2 mm, 4 mm, 6 mm, and 7 mm, respectively, and two identical parallel specimens were prepared for each group of specimens. Subsequently, impact tests were performed on the two parallel specimens of each group of specimens using impact loads of 50 J and 80 J, respectively. Finally, computed tomography technology was used to detect the specimens after the impact test, and computed tomography damage image data of the damaged areas under different impact energy loads were obtained.

[0086] In another embodiment, four groups of carbon fiber composite laminate specimens of different thicknesses were produced according to a specification of 120 mm × 120 mm. The thicknesses of the four groups of carbon fiber composite laminate specimens were 3 mm, 6 mm, 8 mm, and 9 mm, respectively, and two identical parallel specimens were prepared for each group of specimens. Subsequently, impact tests were performed on the two parallel specimens of each group of specimens in sequence using impact loads of 60 J and 90 J. Finally, computed tomography technology was used to detect the specimens after the impact test, and computed tomography damage image data of the damaged area under different impact energy loads were obtained.

[0087] It is understandable that the specifications, shape, thickness and other dimensions of the sample, as well as the magnitude of the impact load can be adaptively set according to actual conditions and are not limited here.

[0088] In some embodiments, the damage mode of the carbon fiber composite laminate under load impacts of different energy levels includes at least one of matrix cracking, delamination, or fiber breakage.

[0089] In some embodiments, the computed tomography damage image dataset is divided into a single damage pattern dataset and a multiple damage pattern dataset. The image data in the single damage pattern dataset contains only one damage pattern, and the image data in the multiple damage pattern dataset contains two or more damage patterns. According to the characteristics and distribution of the damage patterns, the computed tomography damage pattern image dataset is divided into two categories: a single damage pattern image dataset and a multiple damage pattern image dataset. The single damage pattern dataset is used to train and test the CNN model. By classifying, identifying, and quantitatively analyzing the single damage pattern, the optimal hyperparameters of the model are determined, and the performance and accuracy of the model are verified. Based on the optimal hyperparameter combination of the model, the multiple damage pattern dataset is used to test the CNN model. By classifying, identifying, and quantitatively analyzing multiple damage patterns, the robustness and generalization performance of the model are verified.

[0090] In some embodiments, the step of adjusting the mean and standard deviation of the computed tomography lesion pattern image dataset for standardization further comprises:

[0091] T021, transform each computed tomography damage pattern image data into a normal distribution with a mean of 1 and a variance of 0.

[0092] For example, in one embodiment, the mean vector corresponding to the mean of the three RGB channels in a computed tomography image is (0.485, 0.456, 0.406), and the standard deviation vector corresponding to the standard deviation of the three RGB channels is (0.229, 0.224, 0.225). This transforms each CT image sample into a normal distribution with a mean of 1 and a variance of 0.

[0093] In another embodiment, the mean vector corresponding to the means of the three RGB channels in the computed tomography image is (0.5, 0.5, 0.5), and the standard deviation vector corresponding to the standard deviations of the three RGB channels is (0.5, 0.5, 0.5), thereby converting each CT image sample into a normal distribution with a mean of 1 and a variance of 0.

[0094] In some embodiments, the convolutional layer is used to perform damage edge detection and feature extraction on the computed tomography damage image, and the fully connected layer is used to highly abstract and integrate the features extracted by the convolutional layer to output the final laminate damage pattern recognition result;

[0095] Among them, there are 5 convolutional layers and 3 fully connected layers; the network structure of the convolutional neural network model also includes 3 pooling layers and 1 normalized exponential function layer.

[0096] In some embodiments, the number of channels of the first convolutional layer to the fifth convolutional layer of the convolutional neural network model is 16, 32, 64, 96 and 128 respectively.

[0097] In some embodiments, the learning rate of the convolutional neural network model adopts an adaptive adjustment strategy, with an initial learning rate of 1e-4 and a decay rate of 0.1. It changes every 10 epochs, so that the model can adapt to the distribution characteristics of each damage pattern in the target domain more quickly.

[0098] In some embodiments, the step of training and testing the established convolutional neural network model using the computed tomography damage pattern image dataset includes:

[0099] T041, 80 samples of each of the three types of damage patterns in the standardized computed tomography single damage pattern image dataset were randomly selected to form the test set, and the remaining computed tomography single damage pattern images were randomly divided into training and validation sets in a ratio of 4:1;

[0100] T042, train the convolutional neural network model using the training set and validation set, and adjust the hyperparameter settings of the network model during training;

[0101] T043, after model training is completed, the test set is used to test the damage pattern recognition performance of the convolutional neural network model on computed tomography images.

[0102] The training set is used to estimate model parameters, the validation set is used to adjust model hyperparameters and preliminarily evaluate the model's ability to recognize different damage patterns, and the test set is used to finally evaluate the model's performance to ensure that the model can generalize to unseen data.

[0103] In some embodiments, a class activation map is introduced to reweight the weights of the last convolutional layer of a trained convolutional neural network model, visualizing the importance of each input feature, visualizing the decision-making process of the proposed convolutional neural network model, and explaining the reasons behind the model's decisions. These techniques improve the interpretability of the model and allow us to understand the model's decision-making process and important features.

[0104] In some embodiments, the class activation map includes at least one of Grad-CAM, Grad-CAM++, Score-CAM, XGrad-CAM, Ablation-CAM, EigenGrad-CAM, and LayerCAM.

[0105] In some embodiments, Shapley additive explanatory value analysis is introduced for the trained convolutional neural network model to calculate the contribution of different damage pattern features to the prediction results of the convolutional neural network model, provide an explanatory score for each damage pattern feature, and reveal the impact of different damage patterns on the model output.

[0106] In some embodiments, a trained convolutional neural network model was randomly assigned to three damage pattern images from the test set: matrix cracking, delamination, and fiber breakage. Images were rotated 90°, 180°, and 270° clockwise, respectively. Class activation maps and Shapley additive explanatory value analysis were used to visualize the model's recognition and prediction results for the three damaged pattern features after rotation. The images were rotated 90°, 180°, and 270° to test whether the model could still accurately identify the damage pattern under different orientations. The model's recognition ability for the rotated images was further visualized by combining class activation maps and Shapley additive explanatory value analysis.

[0107] In some embodiments, the hyperparameters and weight values ​​of each layer of the optimal convolutional neural network model are stored, and the network model is called to identify a computed tomography multiple damage pattern data set; the model is applied to actual engineering to identify different damage modes of carbon fiber reinforced composite laminates under low-speed impact.

[0108] In some embodiments, multiple damage pattern images are efficiently decoupled and quantitatively analyzed, and interpretable tools such as class activation maps and Shapley additive explanatory values ​​are introduced for transparent interpretation, revealing the potential variables and mechanisms that affect the model prediction results, and visualizing the contribution of each type of damage pattern to the final prediction results; confidence values ​​are introduced to quantitatively analyze the prediction results.

[0109] The following describes in detail a method for identifying damage patterns of carbon fiber reinforced composite materials using an interpretable convolutional neural network through several specific examples.

[0110] Example 1:

[0111] This embodiment provides a method for identifying damage patterns of carbon fiber reinforced composite materials using an interpretable convolutional neural network, comprising the following steps:

[0112] P1, carbon fiber reinforced composite laminate specimens were prepared using the same process parameters: The polymer matrix used in the carbon fiber reinforced composite laminates was a thermosetting epoxy vinyl resin. The resin was cured at room temperature with a curing agent in a mass ratio of 100:2:0.2 with the accelerator and hardener being dimethylaniline and methyl ketone peroxide (MEKP), purchased from The Dow Chemical Company. The reinforcement was a plain-woven polyacrylic acid (PAN)-derived carbon fiber cloth (surface mass density, 300 g / m²) with a single fiber diameter of 10 μm and a 0.2 mm gap between the warp and filler yarns. The carbon fiber fabric was purchased from Xi'an Xiangda Aviation Materials Co., Ltd., China. Conductive silver glue was purchased from Huiteng Company, China, primarily for connecting the electrodes and wires in the carbon fiber reinforced composite laminates. The glue was cured by heating at 100°C for 2 h.

[0113] P2, low-velocity impact testing of the prepared carbon fiber reinforced composite laminates: The carbon fiber reinforced composite laminates were prepared using a vacuum-assisted resin infusion (VARI) process. The length and width of the laminates were 500×500 mm, with thicknesses of 2 mm, 4 mm, 6 mm, and 7 mm, respectively. The carbon fiber reinforced composite laminates were cut into 100 (W) × 100 (L) mm sections using a water jet cutting process. Low-velocity impact testing was performed on the prepared carbon fiber composite laminates using a computer-driven Instron / 9250HV low-velocity drop hammer tester. The hammer mass ranged from 2 to 70 kg, and the impact energy ranged from 0.59 to 1800 J. The position measurement accuracy was ±0.02 mm, and the velocity measurement accuracy was ±0.25%. The incident kinetic energy in the impact tests was set to 50 J and 80 J according to Newton's law.

[0114] P3, Acquiring and Organizing a Low-Velocity Impact Damage Dataset for Carbon Fiber Reinforced Composite Laminate Specimens: All laminate specimens damaged after impact testing were scanned using a YXLON FF35 instrument, along both the thickness and width of the specimens. The laminate impact damage image dataset was established and organized using CT scans. The image data was resized to 227×227 pixels, and the mean and standard deviation of the image data were adjusted for standardization. The mean vector corresponding to the mean of the three RGB channels in the CT scan image is (0.485, 0.456, 0.406), and the standard deviation vector corresponding to the standard deviation of the three RGB channels is (0.229, 0.224, 0.225). The damage modes of carbon fiber reinforced composite laminate specimens are mainly matrix cracking, delamination, and fiber fracture. According to the distribution of damage modes on the image data, the computed tomography (CT) scan image dataset is divided into a single damage mode dataset and a multiple damage mode dataset. In the single damage mode dataset, the image data only contains a single damage mode (matrix cracking, delamination, and fiber fracture), while in the multiple damage mode dataset, the image data contains two or more damage modes.

[0115] P4, Define an Interpretable Convolutional Neural Network Model: A convolutional neural network model for damage pattern recognition in carbon fiber reinforced plastic (CFRP) laminates was constructed. The model consists of two parts: convolutional blocks and fully connected layers. The convolutional blocks detect damage edges and extract features from computed tomography (CT) scan images. The fully connected layers abstract and integrate the features extracted by the convolutional layers to output the final laminate damage pattern recognition results. The convolutional neural network model's architecture includes five convolutional layers, three pooling layers, three fully connected layers, and one softmax layer. The Reluctant Unit (ReLU) activation function is used, and dropout is used to control the model complexity of the fully connected layers. The number of channels in the first to fifth convolutional layers is 16, 32, 64, 96, and 128, respectively. The learning rate of the convolutional neural network model uses an adaptive learning rate strategy, starting at 1e-4 and decreasing to 10% every 10 epochs.

[0116] P5, Training and Testing an Interpretable Convolutional Neural Network Model: A single damage pattern dataset of computed tomography (CT) scan images of damaged CFRP specimens was divided into training, validation, and test sets. The test set consisted of 80 randomly selected CT scan images of each of the three damage types (matrix cracking, delamination, and fiber fracture) in the single damage pattern dataset. The training and validation sets consisted of the remaining image data randomly divided in an 8:2 ratio. A multi-damage pattern dataset of CT scan images of damaged CFRP specimens was used to verify the robustness of the convolutional neural network model. Labels were added to the image data for different damage patterns, with matrix cracking damage labeled as 0, delamination damage labeled as 1, and fiber fracture damage labeled as 2. The convolutional neural network model was trained and validated using the single damage pattern training and validation sets. Hyperparameters of the network model were adjusted during training. After model training, the model's damage pattern recognition accuracy for CT scan images was tested using the test set.

[0117] P6, introduces class activation map (CAM) technology and Shapley additive explanation (SHAP) value interpretable tools to transparently explain the convolutional neural network model: For the trained convolutional neural network model, the class activation map (CAM) technology is introduced to reweight the weights of the last convolutional layer of the convolutional neural network model, visualize the importance of each input feature, and visualize the decision-making process of the proposed convolutional neural network model. The introduced class activation map (CAM) technologies include Grad-CAM, Grad-CAM++, Score-CAM, XGrad-CAM, Ablation-CAM, EigenGrad-CAM, and LayerCAM. For the trained convolutional neural network model, the Shapley additive explanation (SHAP) value analysis is introduced to calculate the contribution of different damage pattern features to the prediction results of the convolutional neural network model, provide an interpretable score for each damage pattern feature, and reveal the impact of different damage patterns on the model output. For the trained convolutional neural network model, three damage pattern images of matrix cracking, delamination, and fiber fracture were randomly selected from the test set and rotated 90°, 180°, and 270° clockwise in sequence. The class activation map (CAM) technology and Shapley additive interpretation (SHAP) value were used to analyze the recognition and prediction results of the three damage pattern features after rotation by the visualization model.

[0118] P7, Generalized Application of Convolutional Neural Network Models: This model stores the hyperparameters and weight values ​​of each layer of a trained interpretable convolutional neural network model. This model is then generalized and applied to identify multiple damage pattern datasets from computed tomography (CT) scans. This model efficiently decouples and quantitatively analyzes multiple damage pattern images, achieving a recognition accuracy of 92.99% for these datasets. The model introduces interpretable tools such as Class Activation Map (CAM) technology and Shapley Additive Explanation (SHAP) values ​​for transparent interpretation, revealing the underlying variables and mechanisms influencing the model's predictions and visualizing the contribution of each damage pattern to the final prediction. Furthermore, confidence levels are introduced for quantitative analysis of the predictions.

[0119] Experimental results

[0120] 1. Figure 4 The present invention proposes an example of an interpretable convolutional neural network model for the output of a single and multiple damage pattern dataset of computed tomography (CT). Figure 4 (a) shows the loss function and accuracy function of the model for single damage pattern recognition during training. It can be seen that the damage function decreases exponentially with the network training process and finally decreases to 0, while the accuracy increases exponentially with the network training process and finally increases to 100%. Figure 4 (b) is the confusion matrix of the damage pattern recognition output results of the convolutional neural network model for the test set of single damage and multiple damage patterns. It can explain that the convolutional neural network has a recognition accuracy of 100% for single damage patterns and a recognition accuracy of 92.99% for multiple damage patterns, realizing efficient decoupling and precise quantification of multiple damage patterns under service conditions.

[0121] 2. Figure 5 This is a schematic diagram of a convolutional neural network model explaining the Class Activation Map (CAM) technology of the present invention. The CAM technology provides a useful local evaluation index for CFRPs damage pattern recognition. Figure 5 The visualization results of different CAM techniques for single-damage pattern image data are shown. Different damage patterns produce different visualization effects, but the highlights remain consistent, concentrating on the damaged area. These highlights are the areas of focus for the network model, explaining the decision-making process of the convolutional neural network model. Due to the rectified linear unit activation function, the CAM heatmap can only produce positive values.

[0122] 3. Figure 6This is a schematic diagram of the Shapley additive interpretation (SHAP) value analysis and interpretation of the convolutional neural network model for single damage pattern recognition results of the example of the present invention. From left to right, the original damage pattern image, the SHAP value for predicting the matrix cracking damage pattern, the SHAP value for predicting the delamination damage pattern, and the SHAP value for predicting the fiber breakage damage pattern are shown. For a given label and a single damage pattern test sample, a positive SHAP value in red indicates that the feature contributes positively to the output label, and a negative SHAP value in blue indicates that the feature contributes negatively to the output label. Figure 6 It can be observed that the red color (positive SHAP value) in the diagonal image is more concentrated. The red area in the prediction results of the matrix cracking damage mode is concentrated in the first image, the red area in the prediction results of the delamination damage mode is concentrated in the second image, and the red area in the prediction results of the fiber rupture damage mode is concentrated in the third image. That is, the classification results of the convolutional neural network are consistent with the true labels. The SHAP value analysis provides an explanatory score for each damage pattern feature. The pattern recognition result is correct, revealing the impact of different damage modes on the model output.

[0123] In summary, the beneficial effects of the present invention are:

[0124] 1. The constructed convolutional neural network model comprehensively considers the importance of convolutional layers and channels. It contains five convolutional layers and has a simple and lightweight structure. The network model uses a larger number of channels in the early stages to simultaneously process more input data information, extract more features of different damage patterns in composite materials, and improve the accuracy of damage pattern recognition tasks. The network model uses a smaller number of channels in the later stages and introduces dropout in the fully connected layers to effectively prevent overfitting of the convolutional neural network model.

[0125] 2. Introducing interpretable tools such as Class Activation Map (CAM) technology and Shapley Additive Explanation (SHAP) values ​​to transparently explain convolutional neural networks. CAM provides local interpretation of convolutional neural networks, effectively demonstrating the reliability of convolutional neural network models in black-box predictions of regions of interest for different damage patterns. SHAP calculates the marginal contribution of features of different damage patterns to the model's predictions, revealing the underlying variables and mechanisms that influence the model's predictions and truly building trust between humans and models.

[0126] 3. The convolutional neural network model is trained and tested using a single damage dataset from computed tomography (CT) scans. The hyperparameters and weight values ​​of each layer of the trained convolutional neural network model are stored. This model can be applied to multiple damage pattern datasets and other damage pattern recognition tasks for carbon fiber reinforced composite materials under impact loads, demonstrating strong robustness.

[0127] The above description is only a partial implementation of the embodiments of the present invention and does not constitute any form of limitation to the application. The protection scope of the embodiments of the present invention is not limited thereto. Any simple modifications, equivalent changes and modifications that can be easily thought of by any technician familiar with this technical field within the technical scope disclosed in the embodiments of the present invention should be covered within the protection scope of the embodiments of the present invention.

Claims

1. A method for identifying damage patterns of carbon fiber reinforced composite materials using an interpretable convolutional neural network, characterized in that: The following steps are involved: Establish a database to collect and organize CT images of carbon fiber composite laminates damaged under low-speed impact loads and establish a CT damage pattern image dataset; Data preprocessing: resizing the image size to a preset size and adjusting the mean and standard deviation of the computed tomography damage pattern image dataset for standardization; Constructing a convolutional neural network model, wherein the convolutional neural network model includes a convolutional layer and a fully connected layer; Training and testing convolutional neural network models, using computed tomography damage pattern image datasets to train and test the established convolutional neural network models, and performing classification, recognition, and quantitative analysis on carbon fiber composite material impact damage pattern image datasets; Interpretation of convolutional neural network models: introducing interpretable tools such as class activation maps and Shapley additive explanatory values ​​to transparently explain convolutional neural network models, reveal the underlying variables and mechanisms that influence model predictions, visualize the contribution of each type of damage pattern to the final prediction results, and perform rotational validation on a computed tomography damage image dataset. Generalization application of the convolutional neural network model. After the model training is completed, the model is generalized and applied to identify multiple damage pattern images and composite material impact damage patterns in actual engineering to verify the robustness of the model.

2. The method according to claim 1, wherein The method further comprises: Composite material preparation steps: - Using carbon fiber or carbon fiber fabric as reinforcement and resin as matrix; -Preparing carbon fiber composite laminates by a lamination process; Sample preparation steps: -Select carbon fiber composite laminates as sample plates; -Manufacture multiple sets of specimens according to predetermined specifications, where the shapes and widths of the specimens in different sets are consistent, but the thicknesses vary; - Prepare two identical parallel samples for each set of specimens; Impact test steps: -Choice of two shock sources with different energy levels; - Two parallel samples of each group of specimens are placed under impact sources of different energies for impact tests; Data processing steps: - Use computed tomography to examine the specimens after impact testing; - Obtain computed tomography damage image data of the damaged area under different impact energy loads.

3. The method according to claim 2, wherein The damage modes of carbon fiber composite laminates under load impacts of different energy levels include at least one of matrix cracking, delamination or fiber breakage.

4. The method according to claim 3, wherein Computed tomography damage image datasets are divided into single damage pattern datasets and multiple damage pattern datasets. The image data in the single damage pattern dataset contains only one damage pattern, while the image data in the multiple damage pattern dataset contains two or more damage patterns.

5. The method according to claim 1, wherein The step of adjusting the mean and standard deviation of the computed tomography damage pattern image dataset for standardization further comprises: Each computed tomography damage pattern image data was transformed into a normal distribution with a mean of 1 and a variance of 0.

6. The method according to claim 1, wherein The convolution layer is used to perform damage edge detection and feature extraction on the CT damage image, and the fully connected layer is used to highly abstract and integrate the features extracted by the convolution layer to output the final laminate damage pattern recognition result; Among them, there are 5 convolutional layers and 3 fully connected layers; the network structure of the convolutional neural network model also includes 3 pooling layers and 1 normalized exponential function layer.

7. The method according to claim 6, wherein The number of channels from the first to the fifth convolutional layer of the convolutional neural network model are 16, 32, 64, 96, and 128, respectively.

8. The method according to claim 6, wherein The learning rate of the convolutional neural network model adopts an adaptive adjustment strategy, with an initial learning rate of 1e-4 and a decay rate of 0.

1. It changes every 10 epochs, allowing the model to adapt more quickly to the distribution characteristics of various damage patterns in the target domain.

9. The method according to claim 4, wherein The steps of training and testing the established convolutional neural network model using the computed tomography damage pattern image dataset include: randomly selecting 80 samples of each of the three types of damage patterns in the standardized computed tomography single damage pattern image dataset to form a test set, and randomly dividing the remaining computed tomography single damage pattern images into a training set and a validation set in a ratio of 4:1; using the training set and the validation set to train the convolutional neural network model, and adjusting the hyperparameter settings of the network model during the training process; after the model training is completed, using the test set to test the damage pattern recognition performance of the convolutional neural network model for the computed tomography images.

10. The method according to claim 1, wherein For the trained convolutional neural network model, a class activation map is introduced to reweight the weights of the last convolutional layer of the convolutional neural network model, visualize the importance of each input feature, visualize the decision-making process of the proposed convolutional neural network model, and explain the reasons behind the model's decision.

11. The method according to claim 10, wherein The class activation map includes at least one of Grad-CAM, Grad-CAM++, Score-CAM, XGrad-CAM, Ablation-CAM, EigenGrad-CAM, and LayerCAM.

12. The method according to claim 1, wherein For the trained convolutional neural network model, Shapley additive explanatory value analysis is introduced to calculate the contribution of different damage pattern features to the prediction results of the convolutional neural network model, provide an explanatory score for each damage pattern feature, and reveal the impact of different damage patterns on the model output.

13. The method according to claim 1, wherein For the trained convolutional neural network model, the three damage pattern images of matrix cracking, delamination, and fiber fracture in the test set were randomly assigned to rotate 90°, 180°, and 270° clockwise in sequence. The class activation map and Shapley additive explanatory value were used to analyze the recognition and prediction results of the three damage pattern features after rotation by the visualization model.

14. The method according to claim 1, wherein The hyperparameters and weight values ​​of each layer of the optimal convolutional neural network model are stored, and the network model is called to identify the multi-damage pattern dataset of computed tomography. The model is applied to actual engineering to identify different damage modes of carbon fiber reinforced composite laminates under low-speed impact.

15. The method according to claim 14, wherein Efficiently decouple and quantitatively analyze images with multiple damage patterns. Interpretable tools such as class activation maps and Shapley additive explanatory values ​​are introduced for transparent interpretation, revealing the underlying variables and mechanisms that influence model predictions and visualizing the contribution of each type of damage pattern to the final prediction. Confidence values ​​are introduced for quantitative analysis of prediction results.

Images