Early warning method and system for crack initiation of compressive apparent crack of rock sample

Abstract

The invention relates to the technical field of rock sample crack data processing, and discloses a rock sample compression apparent crack initiation early warning method and system. The method comprises the following steps: acquiring a historical rock sample surface crack-free image and a crack image to construct a data set, dividing the data set into a training set and a test set, training a convolutional neural network classification model by using the training set, and constructing an apparent crack recognition model; collecting all frames of each video of the microscopic surface image of the rock sample by utilizing high-definition macro photography, inputting each frame of image into an apparent crack recognition model, and recognizing an apparent crack to generate an extended image; and determining a point on the stress-strain curve according to the time corresponding to the first apparent crack image, the point being the crack initiation point of the apparent crack, and determining the appearance stage of the compressive apparent crack of the rock sample and the corresponding crack initiation stress. The system comprises a model construction module, an image recognition module and a crack determination module. The method plays an important role in improving the accuracy, real-time performance and early warning capability of crack initiation detection.

Description

Technical Field

[0001] The present invention relates to the technical field of rock sample crack data processing, and in particular to a method and system for early warning of apparent crack initiation in a rock sample under pressure. Background Art

[0002] Rock specimens are blocks or small pieces of minerals collected from natural rock masses. They are usually used for laboratory research and testing to evaluate the physical and mechanical properties of rocks. Rock specimens play an important role in geological engineering and rock mechanics research. They can be used to conduct a series of experiments and tests to analyze the behavior and characteristics of rocks under different conditions. The deformation stages of rock specimens under uniaxial compression loads include the compaction stage of void cracks, the stable development stage from elastic deformation to microelastic cracks, the unstable fracture development stage, and the post-fracture stage. The accurate identification of the stage of apparent cracks under pressure in rock specimens and the corresponding crack initiation stress can reveal the cracking mechanism and crack evolution process, which is helpful for the study of rock mechanical properties and failure mechanisms. The identification methods of the stage of apparent cracks under pressure in rock specimens and the corresponding crack initiation stress mainly include naked eye observation method, digital image correlation technology, etc. The naked eye observation method is simple and easy, but it is limited by the resolution of the naked eye and the subjectivity of the observer, and cannot accurately identify the stage of apparent cracks under pressure in rock specimens and the corresponding crack initiation stress. Digital image correlation technology measures the displacement and strain field of the rock sample surface by comparing the digital images before and after the rock sample is compressed, thereby monitoring the formation and expansion of cracks. When the rock surface is rough, the application of digital image correlation technology is limited. On the other hand, when using a camera to shoot the surface of the rock sample, the magnification is small, and it is difficult to achieve the identification of the apparent crack initiation stage and the corresponding crack initiation stress of the rock sample under compression on a microscopic scale. In view of the problems existing in the current identification methods of the apparent crack initiation stage and the corresponding crack initiation stress of the rock sample under compression, it is necessary to identify the appearance of apparent cracks and the corresponding crack initiation stress by shooting the more microscopic surface of the rock sample.

[0003] Prior art 1, application number CN202111295467.X discloses a method for real-time monitoring of crack evolution of loaded rock using ultrasound, the method is to perform acoustic wave detection on the loaded rock during uniaxial compression, extract the first wave acoustic time t from the acoustic wave signal time domain diagram of all loaded rock samples, calculate the wave velocity v under each level of load according to the formula, and calculate the wave velocity non-damage ratio α of the rock sample under each level of load through the formula, the smaller the α value, the more cracks. Although the longitudinal wave velocity of the rock sample under different loads is calculated from the time domain diagram, and the non-damage amount of the rock sample wave velocity is defined, the crack changes of the rock sample during loading are quantitatively judged. However, when the rock surface is rough, the relevant applications of this technology will be limited, and the accuracy of crack initiation stress identification cannot be effectively improved.

[0004] Prior art 2, application number CN202111304758.0 discloses a detection method for damage to loaded rock, the method is to simultaneously perform acoustic wave testing and uniaxial compression testing on the rock material, perform data processing on the obtained acoustic wave signal, obtain the acoustic wave signal frequency domain diagram of the rock sample under different loads, divide the frequency interval in the acoustic wave signal frequency domain diagram into multiple equal-width frequency bands, select the frequency band with concentrated acoustic wave signals and obvious peak values ​​in all the divided equal-width frequency band intervals, calculate the frequency domain energy Ef of the rock sample acoustic wave signal in the selected interval, and bring the frequency domain energy Ef into the formula to calculate the frequency domain energy ratio δ of the rock sample acoustic wave signal. The larger the δ value, the more cracks there are and the greater the damage. Although the received acoustic wave signal is processed, the damage of the measured rock sample is judged by defining the dimensionless parameter frequency domain energy ratio δ. However, when the rock surface is rough, the application will be limited, and it is difficult to realize the identification of the apparent crack initiation stage and the corresponding cracking stress of the rock sample under compression on a microscopic scale, which affects the identification effect to a certain extent.

[0005] Prior art three, application number CN202211096278.4 discloses an optical test method for dynamic fracture toughness of straight-cut groove semi-disc rock sample, which belongs to the technical field of dynamic fracture toughness test of rock materials; the luminous intensity of the stress luminescent film under different stresses is measured, the luminous intensity-stress curve is drawn, and the luminous intensity-stress function expression is obtained after fitting; the stress luminescent film is adhered to the surface of the straight-cut groove semi-disc rock sample; the dynamic stress balance is checked and adjusted to coincide with the stress of the incident rod end face and the stress of the reflection rod end face; the rock sample is clamped between the incident rod and the projection rod, and the prefabricated crack of the specimen is parallel to the support rod and located at the center of the two support rods; the cylinder pressure is adjusted, the impact rod is launched, and the rock sample is broken after being impacted, and the luminous intensity of the stress luminescent film is measured; according to the luminous intensity-stress function expression and the measured luminous intensity, the maximum loading force received by the rock sample is calculated, and the stress intensity factor corresponding to the maximum loading force is solved, which is the dynamic fracture toughness of the rock sample. Although the maximum loading force received by the rock sample is calculated, the dynamic fracture toughness of the rock sample is obtained. However, the method is relatively complicated, which increases the operation time and reduces the efficiency of the test.

[0006] At present, the existing technologies 1, 2 and 3 have the problem that when the rock surface is rough, the application of digital image correlation technology will be limited, and it is difficult to realize the identification of the initiation stage of apparent cracks under pressure in rock samples and the corresponding crack initiation stress on a microscopic scale. Therefore, the present invention provides a method and system for early warning of apparent cracks under pressure in rock samples, which identifies the occurrence of cracks and the corresponding crack initiation stress by taking high-definition macro shots of the more microscopic surface of the rock sample. Summary of the invention

[0007] The main purpose of the present invention is to provide a method and system for early warning of apparent crack initiation in rock samples under pressure, so as to solve the problems in the prior art that naked eye observation is limited by the resolution of the naked eye and the initiative of the observer, digital image correlation technology is limited by the rough surface of the rock, and the camera magnification is small, making it difficult to achieve micro-scale observation.

[0008] To achieve the above object, the present invention provides the following technical solutions:

[0009] A method for early warning of apparent crack initiation of a rock sample under pressure, the method comprising:

[0010] Obtain historical rock sample surface images with and without cracks to construct a data set, divide the data set into a training set and a test set, use the training set to train a convolutional neural network classification model, and construct an apparent crack recognition model;

[0011] All frames of each video of the rock sample microscopic surface image are collected using high-definition macro photography, and each frame of the image is input into the apparent crack recognition model to identify the apparent crack and generate an extended image;

[0012] According to the time corresponding to the first apparent crack image, a point on the stress-strain curve is determined, which is the apparent crack initiation point, and the stage of apparent crack appearance and the corresponding crack initiation stress of the rock sample under pressure are determined.

[0013] As a further improvement of the present invention, an apparent crack identification model is constructed, including:

[0014] From the digital image database, we collected labeled images of rock samples with and without cracks on their surfaces, and divided the images into training sets and test sets according to the ratio of 8:2.

[0015] Create convolutional layers, activation layers, pooling layers, fully connected layers, and output layers in the convolutional neural network. Create a flattening layer in the fully connected layer to convert the three-dimensional output into a one-dimensional vector.

[0016] Select the cross entropy loss function, optimizer SGD and evaluation indicators, compile the apparent crack recognition model; use the image data in the training set for training, adjust the model parameters of the apparent crack recognition model, use the test set to evaluate the trained apparent crack recognition model, calculate the accuracy, recall rate and F1 score evaluation indicators of the apparent crack recognition model, and decide whether to fine-tune or optimize the apparent crack recognition model based on the evaluation results.

[0017] As a further improvement of the present invention, compiling an apparent crack identification model includes:

[0018] Import the TensorFlow deep learning library, use the sequential model to build the entire convolutional neural network model structure, and list the specific parameters of each layer;

[0019] Use the cross entropy loss function to create comparison examples; use stochastic gradient descent to add momentum to accelerate convergence and avoid local optimal solutions. Start with SGD and set a small learning rate. After obtaining preliminary results, adjust according to the performance of the training process; select evaluation indicators such as accuracy, recall, precision, and F1 score;

[0020] Combine the model with the selected loss function, optimizer, and evaluation metric to prepare it for training; when compiling, make sure to save all relevant settings of the model; use the compiled model to train by calling the training function to pass in the training data, set the hyperparameters, and verify the model performance after each round of training.

[0021] As a further improvement of the present invention, all frames of each video of the microscopic surface image of the rock sample are collected by high-definition macro photography, including:

[0022] Ensure that the test bench and loading device are in a horizontal state, and calibrate the alignment of the camera, macro lens and loading device; place the rock sample to be tested in the middle position of the loading device; check whether the rock sample is placed horizontally;

[0023] Use multiple camera fixing brackets to fix multiple cameras in the horizontal direction of the rock sample; install a macro lens on each camera, and adjust the shooting angle and focal length of each camera; start the recording software on each camera and configure it, including resolution and frame rate;

[0024] Before starting to apply pressure, confirm that all cameras are in normal working condition and start recording the video of the rock sample under pressure; start the loading device and gradually apply compression force to the rock sample. Set a fixed loading rate as needed; during the process of applying pressure, observe the video captured by the camera and pay close attention to whether cracks appear on the surface of the rock sample; record the video of the entire compression process through the camera to capture the cracks and deformation on the surface of the rock sample; save the recorded video data;

[0025] The collected video data is analyzed frame by frame to capture the occurrence time, location and development process of the rock sample cracks; after confirming that the experimental objectives have been achieved, the loading device is stopped and the rock sample is removed from the loading device; the camera and related equipment are turned off, and the equipment is cleaned and organized for the next experiment.

[0026] As a further improvement of the present invention, gradually applying compressive force to the rock sample comprises:

[0027] Connect and turn on the power of the loading device and accept the control command; the loading device begins to gradually apply compression force to the rock sample according to the preset rate; monitor the display screen of the loading device to observe the change of applied pressure, displacement, and force increase;

[0028] The loading rate was set to a constant 0.5 MPa / s, and the applied force and time information were recorded while the pressure was applied;

[0029] When the set conditions are reached, stop the loading device and record the process of slowing down the pressure to zero; after confirming that it has ended, remove the applied compression force.

[0030] As a further improvement of the present invention, the synchronous operation of the camera's recording software and the loading device includes:

[0031] Install the data acquisition software in the camera's recording software and configure the acquisition rate and required parameters to ensure that each sample has a timestamp;

[0032] Design a trigger circuit. When the loading device starts to apply pressure, use the pressure sensor to close the trigger switch. Through control logic programming, use TTL signal and set it to high level as the trigger timing. When the activation signal is sent, a signal is sent to the camera's recording start port. The camera supports external triggering via LANC, USB or wireless connection.

[0033] Start the recording software and confirm to start recording, and the force of the loading device will be automatically applied to the rock sample, triggering the device to start camera recording.

[0034] As a further improvement of the present invention, determining the initiation point of the apparent crack comprises:

[0035] From the first image where the apparent crack is identified, the timestamp of the frame image is recorded. The timestamp is generated during video recording and is synchronized with the data acquisition system.

[0036] Use the stress-strain data recorded by the data acquisition system, including stress and strain values ​​that change with time; according to the recorded timestamp, find the stress and strain values ​​corresponding to that time in the stress-strain curve data; through the interpolation method, if the crack time observed in the video does not completely match the time recorded in the data, interpolate the data near the time to find the stress and strain values ​​at that moment;

[0037] Use the data visualization tool MATLAB to draw the stress-strain curve using the updated stress and strain data; mark the crack initiation point found on the stress-strain curve, which is the stress value and corresponding strain value when the crack appears;

[0038] According to the relationship between the position of this point and the overall stress-strain curve, the stage at which the crack occurs is analyzed; whether it is the elastic stage, yield stage or other obvious material response characteristics; at the same time, whether this point occurs after the yield point of the material, the behavior of the material is judged; the crack initiation stress and experimental conditions are recorded, and the mapping and stage characteristics required in the experimental report are generated.

[0039] As a further improvement of the present invention, interpolation is performed on data near the time, including:

[0040] obtaining a timestamp of the recording, the timestamp corresponding to the first identified apparent crack image; extracting time-varying stress and strain data from the data acquisition system, organized in a time series format, recording the stress and strain applied at different time points;

[0041] Check the timeline recorded by the data acquisition system to find a time point close to the crack timestamp; if a directly matching timestamp does not exist in the acquired data, it is necessary to determine two time points within a small range;

[0042] Perform interpolation calculations;

[0043] Record the interpolation results and the stress σ obtained by interpolation c and strain ∈ c , the two values ​​are considered as the stress and strain that should exist when the crack occurs.

[0044] To achieve the above object, the present invention also provides the following technical solutions:

[0045] A rock sample pressure apparent crack initiation early warning system, which is applied to the rock sample pressure apparent crack initiation early warning method, the rock sample pressure apparent crack initiation early warning system comprises:

[0046] The model building module is used to obtain the images of historical rock samples with and without cracks on the surface to build a data set, divide the data set into a training set and a test set, use the training set to train the convolutional neural network classification model, and build an apparent crack recognition model;

[0047] An image recognition module is used to collect all frames of each video of the microscopic surface image of the rock sample using high-definition macro photography, input each frame of the image into the apparent crack recognition model, identify the apparent crack and generate an extended image;

[0048] The crack determination module is used to determine a point on the stress-strain curve according to the time corresponding to the first apparent crack image, which is the apparent crack initiation point, and to determine the stage of the apparent crack appearance and the corresponding crack initiation stress of the rock sample under pressure.

[0049] As a further improvement of the present invention, it also includes: a rock sample, a camera fixing bracket, a camera, a loading device, and a macro lens;

[0050] Among them, the rock sample is placed in the middle of the loading device, and multiple cameras are fixed in the horizontal direction of the rock sample. The cameras are fixed by multiple camera fixing brackets, and the macro lenses are installed on the cameras; the cameras record the video of the entire process of the rock sample being subjected to pressure, and the macro lenses are evenly arranged along the horizontal direction of the cylindrical rock sample and equidistantly arranged along the vertical direction.

[0051] The invention constructs a data set and trains a model, by acquiring crack-free images and cracked images, to construct a complete data set for subsequent model training and testing; divides the data set into a training set and a test set, laying a foundation for model training and performance evaluation; uses the training set data to train a convolutional neural network (CNN) classification model, and constructs an apparent crack recognition model, so that it can recognize and classify crack features in the image. The construction of the data set provides a sample basis required for machine learning, ensuring that the model can learn the correct crack features; by dividing the training set and the test set, the generalization ability of the model can be verified to ensure its performance on unknown data; the trained model can automatically identify cracks on the surface of the rock sample, enhances the efficiency and accuracy of detection, and reduces the dependence on manual detection. Image acquisition and crack recognition, using high-definition macro photography technology to collect microscopic surface images of rock samples under pressure, and extract each frame of the video to form a continuous image sequence; each frame of the image is input into the trained apparent crack recognition model to automatically identify the apparent crack image. Real-time monitoring: Through high-definition macro photography technology, the surface changes of rock samples can be continuously monitored during the compression process, and the dynamic information of crack generation can be captured in time; High-precision identification: The application of the CNN model makes the identification of cracks more accurate, which can effectively reduce the probability of missed detection and false detection; Data accumulation: The images collected in this step provide a large amount of empirical data for subsequent analysis and research, which helps to understand the mechanism and law of crack formation. Determine the crack initiation point and stress-strain curve. According to the time corresponding to the first identified apparent crack image, determine a key point on the stress-strain curve, that is, the crack initiation point; Key parameter identification: Accurately identifying the crack initiation point is crucial for analyzing the mechanical behavior of rock samples during compression, and can provide basic data for subsequent research; Effective early warning: By determining the time of crack occurrence and the corresponding stress value, the compressive limit of the material can be pre-judged, and timely measures can be taken to reduce the risk of accidents; Optimized design: These parameters help to optimize material selection and design schemes in geotechnical engineering and structural design, thereby improving overall safety. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] Figure 1 A schematic diagram of the steps of an embodiment of a method for early warning of apparent crack initiation of a rock sample under pressure according to the present invention;

[0053] Figure 2It is a principle diagram of an embodiment of a method for early warning of apparent crack initiation of a rock sample under pressure of the present invention;

[0054] Figure 3 A schematic flow chart of the steps of constructing an apparent crack recognition model for an embodiment of a method for early warning of apparent crack initiation in rock samples under pressure according to the present invention;

[0055] Figure 4 A schematic flow chart of the steps of collecting all frames of each video of the microscopic surface image of a rock sample by high-definition macro photography in one embodiment of the method for early warning of the initiation of apparent cracks in a rock sample under pressure of the present invention;

[0056] Figure 5 A schematic flow chart of the steps of determining the initiation point of an apparent crack in an embodiment of a method for early warning of the initiation of apparent cracks in a rock sample under pressure according to the present invention;

[0057] Figure 6 This is a schematic diagram of the functional modules of an embodiment of the rock sample compression apparent crack initiation early warning system of the present invention;

[0058] Figure 7 The structure of an embodiment of the rock sample compression apparent crack initiation early warning system of the present invention is schematically shown. Figure 1 ;

[0059] Figure 8 The structure of an embodiment of the rock sample compression apparent crack initiation early warning system of the present invention is schematically shown. Figure 2 ;

[0060] Fig. 9 The structure of an embodiment of the rock sample compression apparent crack initiation early warning system of the present invention is schematically shown. Figure 3 ;

[0061] Fig.10 The structure of an embodiment of the rock sample compression apparent crack initiation early warning system of the present invention is schematically shown. Figure 4 ;

[0062] Fig.11 A schematic diagram of functional modules of an embodiment of an electronic device of the present invention;

[0063] Fig.12 This is a functional module diagram of an embodiment of the storage medium of the present invention. DETAILED DESCRIPTION

[0064] 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.

[0065] The terms "first", "second" and "third" in the present invention are only used for descriptive purposes and cannot be understood as indicating or implying relative importance or implicitly indicating the number of indicated technical features. Therefore, the features defined as "first", "second" and "third" can explicitly or implicitly include at least one of the features. In the description of the present invention, the meaning of "multiple" is at least two, such as two, three, etc., unless otherwise clearly and specifically defined. All directional indications (such as up, down, left, right, front, back...) in the embodiments of the present invention are only used to explain the relative position relationship, movement, etc. between the components under a certain specific posture (as shown in the accompanying drawings). If the specific posture changes, the directional indication also changes accordingly. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device that includes a series of steps or units is not limited to the listed steps or units, but optionally also includes steps or units that are not listed, or optionally also includes other steps or units inherent to these processes, methods, products or devices.

[0066] Reference to an "embodiment" herein means that a particular feature, structure, or characteristic described in conjunction with the embodiment may be included in at least one embodiment of the present invention. The appearance of the phrase in various places in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0067] like Figure 1 As shown, this embodiment provides an embodiment of a method for early warning of apparent crack initiation of a rock sample under pressure. In this embodiment, the method for early warning of apparent crack initiation of a rock sample under pressure specifically includes the following steps:

[0068] Step S1: Acquire historical rock sample surface images without cracks and images with cracks to construct a data set, divide the data set into a training set and a test set, use the training set to train a convolutional neural network classification model, and construct an apparent crack recognition model;

[0069] Step S2: using high-definition macro photography to collect all frames of each video of the rock sample microscopic surface image, inputting each frame of the image into the apparent crack recognition model, identifying the apparent crack and generating an extended image;

[0070] Step S3: Determine a point on the stress-strain curve according to the time corresponding to the first apparent crack image, which is the apparent crack initiation point, and determine the stage of the apparent crack appearance of the rock sample under pressure and the corresponding crack initiation stress.

[0071] Preferably, step S1 of this embodiment constructs a data set and model training, and constructs a complete data set by acquiring crack-free images and cracked images for subsequent model training and testing; divides the data set into a training set and a test set to lay the foundation for model training and performance evaluation; uses the training set data to train a convolutional neural network (CNN) classification model, and constructs an apparent crack recognition model so that it can recognize and classify crack features in the image. The construction of the data set provides the sample basis required for machine learning, ensuring that the model can learn the correct crack features; by dividing the training set and the test set, the generalization ability of the model can be verified to ensure its performance on unknown data; the trained model can automatically identify cracks on the surface of the rock sample, enhance the efficiency and accuracy of detection, and reduce the dependence on manual detection. Step S2 image acquisition and crack recognition, using high-definition macro photography technology to collect microscopic surface images of rock samples under pressure, and extract each frame of the video to form a continuous image sequence; each frame of the image is input into the trained apparent crack recognition model to automatically identify the apparent crack image. Real-time monitoring: Through high-definition macro photography technology, the surface changes of rock samples can be continuously monitored during the compression process, and the dynamic information of crack generation can be captured in time; High-precision identification: The application of the CNN model makes the identification of cracks more accurate, which can effectively reduce the probability of missed detection and false detection; Data accumulation: The images collected in this step provide a large amount of empirical data for subsequent analysis and research, which helps to understand the mechanism and law of crack formation. Step S3 determines the crack initiation point and the stress-strain curve. According to the time corresponding to the first identified apparent crack image, a key point on the stress-strain curve is determined, that is, the crack initiation point; Key parameter identification: Accurate identification of the crack initiation point is crucial for analyzing the mechanical behavior of rock samples during compression, and can provide basic data for subsequent research; Effective early warning: By determining the time when the crack appears and the corresponding stress value, the compressive limit of the material can be pre-judged, and timely measures can be taken to reduce the risk of accidents; Optimized design: These parameters help to optimize material selection and design schemes in the process of geotechnical engineering and structural design, thereby improving overall safety (for specific principles, refer to the attached Figure 2 ).

[0072] In summary, each step of the method for early warning of apparent crack initiation in rock samples under pressure in this embodiment plays an important role in improving the accuracy, real-time performance and early warning capability of crack initiation detection. By making full use of modern image processing and machine learning technologies, crack monitoring in the process of rock samples under pressure is made more systematic and automated, providing solid technical support for the research and practice of geological engineering and materials science.

[0073] Furthermore, if Figure 3 As shown, the process of constructing the apparent crack identification model in step S1 specifically includes the following steps:

[0074] Step S11: Collect annotated images of rock samples with and without cracks on their surfaces from a digital image database, and divide the images into a training set and a test set according to a ratio of 8:2;

[0075] Step S12: Create a convolutional layer, activation layer, pooling layer, fully connected layer and output layer in the convolutional neural network. Create a flattening layer in the fully connected layer to convert the three-dimensional output into a one-dimensional vector. The number of neurons is 512, the activation layer is ReLU, and Dropout is 0.5; the number of neurons in the output layer is 2 (the classification task is no crack and with crack) activation layer, Softmax (multi-category classification, output probability);

[0076] Step S13: Select the cross entropy loss function, optimizer SGD and evaluation index, compile the apparent crack recognition model; use the image data in the training set for training, adjust the model parameters of the apparent crack recognition model, use the test set to evaluate the trained apparent crack recognition model, calculate the accuracy, recall rate and F1 score of the apparent crack recognition model, and decide whether to fine-tune or optimize the apparent crack recognition model according to the evaluation results.

[0077] Preferably, in step S11 of this embodiment, data collection and division, annotated rock sample images are collected from the digital image database, including samples without cracks and with cracks, to provide basic data support for the model, ensuring that the model can learn features of different categories; the images are divided into training sets and test sets in a ratio of 8:2. This division ensures that during the training process, the model can use rich data to learn, while the test set is used to evaluate the generalization ability of the model; enhance the robustness of the model: through a large number of sample training, the model's adaptability to unknown test data can be improved; model evaluation: the existence of the test set allows the performance of the model on unseen data to be objectively evaluated, thereby verifying the generalization ability of the model. Step S12 creates a convolutional neural network architecture and constructs network layers: creates a combination of convolutional layers, activation layers, pooling layers, flattening layers, fully connected layers, and output layers. This hierarchical structure allows the model to extract features from the image and understand complex patterns; convolutional layers: learn the spatial hierarchical features of the image layer by layer through small convolution kernels; activation layers (ReLU): introduce nonlinearity so that the model can learn more complex features; pooling layers: downsampling reduces feature dimensions, accelerates training and prevents overfitting; flattening layers: convert three-dimensional tensors into one-dimensional vectors for input to fully connected layers; fully connected layers: integrate previously extracted features to help the model make final decisions; output layers: provide the final classification probability for distinguishing between crack-free and crack-existing images. Feature extraction: The design of the network layers enables the model to improve its ability to express image features layer by layer, from simple to complex; strong adaptability: through the configuration layer, the network can adapt to features of different complexities to support specific tasks (crack recognition). Step S13: Model compilation, training and evaluation, selection of loss function and optimizer: cross entropy loss function is used for multi-classification tasks, SGD (stochastic gradient descent) optimizer is used to update model parameters; Model training: multiple rounds of training are performed using the training set to optimize the model weights; Model evaluation: accuracy, recall and F1 score are calculated on the test set. Performance evaluation: through these quantitative indicators, we can clearly understand the reliability of the model in the real world; Model optimization decision: evaluation results (such as accuracy and F1 score) can guide the next step of model fine-tuning or improvement to ensure the effectiveness of the model in the production environment; Combination of theory and practice: by scientifically selecting loss functions and optimizers, matching with specific application scenarios, the learning efficiency of the model is improved.

[0078] In summary, the design and execution of each step in this embodiment is intended to ensure that the apparent crack recognition model can effectively learn and accurately identify cracks. It provides a scientific basis for the construction and optimization of the model, so that the final model can not only perform well on the training set, but also show good generalization ability for new data in practical applications; it is an important guarantee for the successful implementation of machine learning and deep learning.

[0079] Furthermore, the process of compiling the apparent crack identification model in step S13 specifically includes the following steps:

[0080] Step S131: Import the TensorFlow deep learning library, use the sequential model to build the entire convolutional neural network model structure, and list the specific parameters of each layer;

[0081] Step S132: Use the cross entropy loss function to create a comparison example; use stochastic gradient descent to add momentum to accelerate convergence and avoid local optimal solutions, start with SGD, and set a small learning rate. After obtaining preliminary results, adjust according to the performance of the training process; select evaluation indicators such as accuracy, recall, precision, and F1 score;

[0082] Step S133: Combine the model with the selected loss function, optimizer, and evaluation metric to prepare it for training; when compiling, ensure that all relevant settings of the model are saved; use the compiled model to pass the training data into the training by calling the training function, set the hyperparameters, and verify the model performance after each round of training.

[0083] Preferably, step S131 of this embodiment constructs a convolutional neural network model structure, imports the TensorFlow library to ensure that the deep learning model can be built using the TensorFlow framework, and utilizes its powerful functions and optimization tools; sequential model construction: by using a sequential model, neural network components (such as convolution layers, activation layers, pooling layers, etc.) are added layer by layer to form a linear stacking structure that develops layer by layer; list the specific parameters of each layer: the parameter settings of each layer (such as the number of convolution kernels, size, stride, activation function, etc.) are listed in detail to help clarify the construction logic of the model. A clear model architecture design makes subsequent debugging and optimization more efficient; understanding the function of each layer can help design a network structure that is more suitable for a specific task; by constructing convolution layers and other layers, the model can extract and learn useful features in the input data, which is the basis for achieving effective pattern recognition. Step S132 selects a loss function, optimizer, and evaluation index. The cross entropy loss function is used to quantify the gap between the probability distribution of the model output and the true label, which is a commonly used loss function in classification tasks. Create comparative examples: Through example comparison, help evaluate and understand the impact of different choices on model performance (such as choosing different loss functions) so that adjustments can be made in real time; Select optimizer: Use stochastic gradient descent (SGD) and its variants (such as adding momentum), which can effectively optimize parameters and accelerate convergence while reducing the risk of falling into local optimal solutions; Set learning rate: Initially setting a small learning rate can effectively avoid oscillations during training and ensure a smooth optimization process; Select multiple evaluation indicators: In addition to accuracy, recall rate, precision rate and F1 score are also introduced to provide multi-angle evaluation of model performance; Select appropriate loss functions and optimizers to ensure that the model can effectively learn and optimize parameters to improve the accuracy of classification tasks; The diversity of evaluation indicators makes the evaluation of model performance more comprehensive, and more important performance indicators can be prioritized in different scenarios, such as recall rate in an environment with high error costs. Step S133: compile the model and train it, combine the model with the selected loss function, optimizer and evaluation index to form a complete training plan, and prepare the model for training; save the relevant parameters and structural settings of the model to ensure that the training process is reproducible and convenient for subsequent adjustment or optimization; use the training function to train the model: specifically call the training function, use the training data to train the model, and set appropriate hyperparameters (such as batch size, number of training rounds, etc.); use the validation set to check the performance of the model after each round of training, and adjust the model or hyperparameters by observing the loss and evaluation indicators. Compiling the model allows the model to execute the training process, which is an important step in model construction. This stage ensures that all components work in coordination and improves the learning effect of the model; by continuously verifying the performance of the model, potential problems can be discovered early and necessary adjustments can be made, thereby improving the accuracy and reliability of the model and ensuring that the expected performance level is finally achieved.

[0084] In summary, this embodiment can better understand the complexity and importance of the model compilation and training process by clarifying the technical effects and objectives of each step; it provides a theoretical basis for optimizing model performance, adjusting hyperparameters, and determining the best training route, ensuring that the model can ultimately perform well in the apparent crack identification task.

[0085] Furthermore, if Figure 4 As shown, the process of collecting all frames of each video of the rock sample microscopic surface image by high-definition macro photography in step S2 specifically includes the following steps:

[0086] Step S21: Ensure that the test bench and the loading device are in a horizontal state, calibrate the alignment of the camera, the macro lens and the loading device; place the rock sample to be tested in the middle position of the loading device; check whether the rock sample is placed horizontally;

[0087] Step S22: Use multiple camera fixing brackets to fix multiple cameras in the horizontal direction of the rock sample; install a macro lens on each camera, and adjust the shooting angle and focal length of each camera; start the recording software on each camera and configure it, including resolution, frame rate, etc.;

[0088] Step S23: before starting to apply pressure, confirm that all cameras are in normal working condition, and start recording the video of the rock sample being pressed; start the loading device, gradually apply the compressive force to the rock sample, and set a fixed loading rate as needed; in the process of applying pressure, observe the video captured by the camera, and pay close attention to whether cracks appear on the surface of the rock sample; record the video of the entire compression process through the camera, and capture the cracks and deformation on the surface of the rock sample; save the recorded video data;

[0089] Step S24: Analyze the collected video data frame by frame to capture the occurrence time, location and development process of the rock sample cracks; after confirming that the experimental goal has been achieved, stop the loading device and take the rock sample out of the loading device; turn off the camera and related equipment, and clean and organize the equipment for use in the next experiment.

[0090] Preferably, in step S21 of this embodiment, the experimental preparation and rock sample placement ensure that the experimental table and the loading device are in a horizontal state to avoid uneven pressure due to unevenness, which affects the experimental results; the alignment of the camera, macro lens and loading device ensures that the rock sample is always in the field of view of the camera during the entire compression process, reducing the blind spot in image capture; the rock sample to be tested is placed in the middle of the loading device, and its horizontal placement is checked to ensure the uniformity of the force; ensuring that the initial setting of the experiment is accurate is crucial for obtaining reliable experimental data. It helps to avoid errors caused by improper placement of experimental devices and rock samples, thereby improving the accuracy and repeatability of data. Step S22 camera and lens setting fixes the camera on a horizontal plane through multiple camera fixing brackets to ensure that the dynamic changes of the rock sample surface are recorded during the compression process; a macro lens is installed on each camera to clearly capture the tiny details of the rock sample surface, especially the appearance of cracks; configure the resolution and frame rate of the camera to ensure that high-quality video and images are obtained for detailed post-analysis; through reasonable camera layout and configuration, high-resolution and high-frame-rate video data can be obtained. This is crucial for observing subtle changes on the surface of rock samples, such as the generation and expansion of cracks, and helps the depth and breadth of subsequent analysis. Step S23 video recording and data collection, confirm the normal working status of all cameras before starting to apply pressure to ensure that data can be continuously obtained during the experiment; start the loading device, gradually apply compression force to the rock sample, apply load at the set rate, and observe the behavior of the rock sample; observe the recording process in real time to ensure that the appearance of cracks on the surface of the rock sample can be captured and recorded as video data; the video recording process is very critical, it provides a complete compression record, including the load changes on the rock sample and the formation and development of surface cracks. These data will be used for subsequent analysis and research to help understand the failure mechanism of rock samples. Step S24: Video data analysis and end of the experiment, analyze the collected video data frame by frame, accurately record the occurrence time, location and development process of the cracks, and extract key data; stop applying pressure after confirming that the goal has been achieved, remove the rock sample from the loading device, and turn off the equipment at the same time for sorting and cleaning; frame-by-frame analysis allows researchers to understand the dynamic evolution of crack occurrence in detail, which is the basis for studying the mechanical properties of geotechnical materials and the development of early warning systems. The conclusion of the experiment and the cleaning of the equipment can keep the laboratory running normally and prepare for subsequent experiments.

[0091] In summary, this embodiment systematically collects microscopic surface images of rock samples through the above four steps, from experimental preparation, equipment configuration to data recording and post-analysis, ensuring that each link maximizes the accuracy and reliability of the data. This is particularly important in the fields of geotechnical engineering and materials science.

[0092] Further, the process of gradually applying the compressive force to the rock sample in step S23 specifically includes the following steps:

[0093] Step S231: connect and turn on the power of the loading device and accept the control command; the loading device begins to gradually apply the compressive force to the rock sample according to the preset rate; monitor the display screen of the loading device to observe the applied pressure change, displacement, force increase, etc.;

[0094] Step S232: The loading rate is set to a constant 0.5 MPa / s, and while applying pressure, the applied force and time information are recorded;

[0095] Step S233: When the set conditions are met, the loading device stops working and the process of slowing down the pressure to zero is recorded; after confirming that it has ended, the applied compressive force is released.

[0096] Preferably, step S231 of this embodiment connects and starts the loading device, ensures that the power connection of the loading device is normal, the display screen functions normally, and all control systems are available; this is the prerequisite for the start of the loading process; accepts control commands: the loading device can receive preset control commands to ensure that it applies pressure as planned; at this time, the display screen of the monitoring loading device provides real-time feedback, displaying key information such as applied pressure and displacement. Real-time monitoring: Continuously observe the changes in applied pressure, displacement and force increase, so as to promptly identify problems and adjust application strategies to ensure the accuracy of data and the safety of the experiment; ensuring that all functions of the loading device are normal before applying pressure is to improve the reliability of the experiment. Real-time monitoring helps to detect potential problems early, avoid experimental accidents, and ensure the safety of scientific researchers. Step S232 sets a constant loading rate, and sets the loading rate to a constant 0.5MPa / s, which can ensure the constancy and controllability of force application during the entire pressure application process; while applying pressure, the applied force and time information are recorded; data recording can use a data acquisition system for use in subsequent analysis and research; the constant loading rate can ensure that the response of the rock sample under stress is predictable and repeatable, thereby providing a reliable data basis for subsequent analysis. Because the data is recorded in real time, the experiment is traceable, which helps subsequent researchers to reproduce the same experimental conditions and environment. Step S233 stops the loading device and unloads the compression force, identifies and determines whether the experimental conditions have reached the preset value (such as the maximum pressure or the appearance of cracks in the rock sample). Once reached, the loading device is stopped immediately to prevent the rock sample from being overloaded and causing abnormal damage or explosion; the process of slowing down from the maximum pressure to zero can be recorded to observe the changes in the rock sample during the pressure relief process, which is also very important for understanding the behavior of the material after the load is removed; after confirming that the experiment is over, the applied compression force is removed to ensure that the rock sample is properly handled to avoid secondary damage caused by unloading too quickly; by controlling the unloading process in a timely manner, the integrity of the rock sample structure is enhanced and the accuracy of the data is ensured. Recording the unloading process helps to study the performance of materials under different load conditions and lays the foundation for materials science and engineering applications.

[0097] In summary, this embodiment uses three steps to ensure the rigor and safety of the experiment in the entire process of gradually applying compression force to the rock sample; each link plays a key role, effectively improving the data quality and the repeatability of the experiment, and providing a good foundation for subsequent crack analysis and material mechanical properties research; this systematic method can effectively reduce unnecessary errors and ensure the reliability of the results obtained in the experiment.

[0098] Furthermore, the synchronous operation process of the camera's recording software and the loading device in step S2 specifically includes the following steps:

[0099] Step S25: Install data acquisition software in the camera's recording software, and configure the acquisition rate (e.g., 100 samples per second) and required parameters (pressure, displacement, etc.), ensuring that each sample has a timestamp.

[0100] Step S26: Design a trigger circuit. When the loading device starts to apply pressure, use the pressure sensor to close the trigger switch. Through control logic programming, use the TTL signal and set it to a high level as the trigger timing. When the activation signal is issued, a signal is sent to the recording start port of the camera. The camera supports external triggering via LANC, USB or wireless connection.

[0101] Step S27: Start the recording software and confirm to start recording, and automatically apply the force of the loading device to the rock sample, triggering the device to start camera recording.

[0102] Preferably, step S25 of this embodiment installs and configures data acquisition software. After installing the data acquisition software, configure the acquisition rate (such as 100 samples per second) to ensure real-time monitoring of applied pressure and displacement changes. This configuration enables each sample in the experiment to accurately record its timestamp information, thereby achieving accurate data management and later analysis; configuring the required parameters (such as pressure, displacement, etc.) can provide comprehensive state monitoring for the experiment, which helps to capture key experimental dynamics; by ensuring that the data collects all key parameters and has a timestamp, it can provide high-quality data records during the experiment and enhance the reliability and accuracy of data analysis. In this way, when studying crack evolution or other processes in the later stage, in-depth analysis can be performed based on accurate time series data. Step S26 designs and programs the trigger circuit, designs the trigger circuit and uses the pressure sensor to construct a closed switch, which can automatically trigger the camera recording, and controls the start of the camera through the TTL signal. This method ensures that the state changes of the rock sample can be captured in real time during the pressure application process; using a high-level signal to trigger the camera can ensure that the camera can record synchronously when the pressure is applied at the beginning, which ensures that the shooting is consistent with the loading time, thereby improving the validity of the data in the subsequent analysis; through the automated trigger mechanism, the lag and human error of the experimenter's manual operation are eliminated, so that the data recording and the pressure application are always synchronized, thereby enhancing the scientific nature of the experiment and the accuracy of the data. Step S27 starts the recording software and applies pressure, starts the camera's recording software and confirms the start of recording, ensuring that all key moments of the experiment (especially the state of the rock sample at the beginning of the pressure) are captured; during the pressure application process, the camera starts recording while the pressure is applied, so that the state and changes of the rock sample during the entire experiment can be recorded in real time and reflect the real-time loading situation; ensuring the synchronization of video recording and loading data, allowing researchers to correspond the specific images of the video with pressure, displacement and other data in the later analysis, so as to further understand the mechanical behavior of the material, especially the performance in the development and expansion of cracks. This multi-dimensional data collection method will greatly improve the reliability of experimental results and the depth of subsequent research.

[0103] In summary, through the implementation of these three steps, the camera recording software and the loading device can achieve efficient and automated synchronous operation, providing high-quality data collection and recording for the experiment. The smooth progress of this process significantly improves the integrity, accuracy and traceability of the rock sample compression process data, thereby providing a solid data foundation for subsequent analysis and research.

[0104] Furthermore, if Figure 5 As shown, the process of determining the starting point of the apparent crack in step S3 specifically includes the following steps:

[0105] Step S31: identify the first image of the apparent crack and record the timestamp of the frame image. The timestamp is generated during video recording and is synchronized with the data acquisition system.

[0106] Step S32: using the stress-strain data recorded by the data acquisition system, including stress and strain values ​​that change with time; searching for stress and strain values ​​corresponding to the time in the stress-strain curve data according to the recorded timestamp; by interpolation, if the crack time observed in the video does not completely match the time of the data recording, interpolating the data near the time to find the stress and strain values ​​at that moment;

[0107] Step S33: Use the data visualization tool MATLAB to draw a stress-strain curve using the updated stress and strain data; mark the found crack initiation point on the stress-strain curve, which is the stress value and corresponding strain value when the crack appears;

[0108] Step S34: Analyze the stage of crack occurrence based on the relationship between the position of the point and the overall stress-strain curve; it can be the elastic stage, yield stage or other obvious material response characteristics; at the same time, determine whether the point occurs after the material yield point to judge the behavior of the material; record the crack initiation stress and experimental conditions, and generate the mapping and stage characteristics required in the experimental report.

[0109] Preferably, step S31 of this embodiment records the crack image and timestamp, and the data recording accuracy: extract the timestamp from the first image of the identified apparent crack to ensure the accuracy of the recorded time information; the timestamp is consistent with the timestamp generated during video recording and synchronized with data acquisition, which can provide a basis for subsequent data association; time series correlation: ensure that the acquired crack image time information can maintain temporal consistency with the data loaded in the experiment (such as stress, strain, etc.), providing a basis for time series data for subsequent analysis; through accurate timestamp recording, lay the foundation for subsequent stress and strain data search, so that the experimental data processing is closely integrated with the moment when the crack appears, which is helpful for in-depth analysis of the crack formation mechanism and material properties. Step S32 searches for stress-strain data, uses the recorded timestamp to search for the corresponding stress and strain values, and synchronizes the video recording with the experimental data; through the interpolation method, ensure that even if there is a slight time deviation, accurate stress and strain values ​​can still be obtained; enhance data analysis capabilities: interpolation calculations make it possible to more accurately connect the moment when the crack occurs with the dynamic changes of stress and strain in time, helping to build a more detailed stress-strain history; through this process, ensure that the simultaneous conditions of crack occurrence can be accurately identified, providing key data for subsequent analysis. This precise matching is crucial to understanding the changes in materials during stress, and helps to determine the limit state and failure mechanism of materials. Step S33 draws a stress-strain curve, uses data visualization tools such as MATLAB, presents stress-strain data in the form of a curve, and intuitively displays the response of the material during the application of pressure; clearly marks the initiation point on the curve, making it easy to identify the key moment of crack occurrence, which is convenient for analysts to conduct subsequent research; through visualization, it is possible to quickly identify important features in the stress-strain curve, intuitively understand the initiation behavior of the crack and the response of the material, and provide important visual basis for subsequent discussions and conclusions. Step S34 analyzes the stage of crack occurrence, determines the behavior of the material when the crack occurs (such as elastic stage, yield stage, etc.) by analyzing the position of the crack initiation point in the stress-strain curve, and judges the response characteristics of the material; records the crack initiation stress, the corresponding strain value and the experimental conditions, and provides necessary information for generating experimental reports and further research; this step lays the foundation for a deep understanding of the mechanical behavior of materials under different pressures, and can help researchers evaluate the performance of materials, failure mechanisms and crack progression. In addition, it has important guiding significance for the design and application of materials, especially in the engineering field.

[0110] In summary, this embodiment provides a systematic methodology for determining the crack initiation point through the implementation of the above steps, from recording the image timestamp to calculating and analyzing the stress-strain data, and then to visualizing and clarifying the phased characteristics. This enables a comprehensive understanding of the behavior of the material during the stress process, improves the understanding of material properties, and provides important data support for subsequent material optimization and crack monitoring work.

[0111] Furthermore, the process of interpolating the data near the time in step S32 specifically includes the following steps:

[0112] Step S321: obtaining a recorded timestamp corresponding to the first identified apparent crack image; extracting stress and strain data varying with time from a data acquisition system, organizing them in the form of a time series, and recording the stress and strain applied at different time points;

[0113] Step S322: Check the time axis recorded by the data acquisition system to find a time point close to the crack timestamp; if a directly matching timestamp does not exist in the acquired data, it is necessary to determine two time points within a small range (one earlier and one later than the crack timestamp);

[0114] Select the upper and lower limit time points: Assume that the timestamp of the crack is (t c ), find two time points (t1) and (t2) in the data, where (t1 <t c <t2);

[0115] Extract the corresponding stress (σ1, σ2) and strain (∈1, ∈2) values ​​from these two time points;

[0116] Step S323: performing interpolation calculation;

[0117] Linear interpolation formula: If you select the upper and lower time points, use the linear interpolation method to estimate the crack time stamp (t c ) when:

[0118]

[0119] Explanation of interpolation calculation: The meaning of the formula is to use the known two-point information (t1, t2) to calculate the stress and strain when the crack appears using the straight line equation; through the proportion, that is, Weigh changes in stress and strain;

[0120] Step S324: Record the interpolation results, and the stress (σ c ) and strain (∈ c ), the two values ​​are considered as the stress and strain that should exist when the crack occurs.

[0121] Preferably, step S321 of this embodiment obtains the recorded timestamp and stress-strain data, time synchronization: first, accurately obtain the timestamp of the crack image to ensure that the time can be synchronized with other data (such as stress and strain) in time; data extraction: extract the stress and strain data that change with time from the data acquisition system to ensure that these data accurately and comprehensively reflect the dynamic response of the material during loading; ensure the accuracy of the preliminary information and provide a solid data basis for subsequent analysis. Accurate timestamps and data records are the prerequisites for effective interpolation and directly affect the reliability and accuracy of the final result. Step S322 checks the data timeline and selects a time point, data matching: checks the time record of the data acquisition system to find two time points close to the crack timestamp; ensures that the selected time point can reflect the specific moment of the crack as accurately as possible; establishes upper and lower limits: determines the range of interpolation through the selected time point, and provides the necessary upper and lower limits for the subsequent estimation of stress and strain values; in this way, the problem of timestamp mismatch can be overcome, ensuring that the most recent stress and strain data can be used for interpolation calculation when the crack appears, thereby improving the accuracy and reliability of the data. Step S323 performs interpolation calculations using linear interpolation: by selecting upper and lower limit time points, the linear interpolation method is used to accurately calculate the stress and strain at the crack timestamp; the linear interpolation formula can clearly convert known data into estimated values ​​at a specific time point, thereby effectively utilizing existing data; information trade-off: interpolation calculations can not only process isolated recording points, but also combine the information of two time points to obtain more accurate interpolation results; through interpolation calculations, stress and strain values ​​can be accurately obtained at the moment of crack formation, which is crucial for understanding and analyzing the failure characteristics of materials; making the analysis of material mechanics more accurate, providing an important basis for subsequent research and application. Step S324 records the interpolation results. Data integration: the stress and strain values ​​calculated by interpolation are recorded. These two values ​​can accurately reflect the material state when the crack occurs. Data utilization: the interpolation results are integrated into the experimental data set to provide timely reference values ​​for subsequent analysis. Recording the interpolation results makes the stress and strain when the crack occurs traceable and analyzable, providing important data support. It provides a basis for further crack dynamic analysis and material behavior research, and provides rich information for generating experimental reports and subsequent improvement of experimental design.

[0122] In summary, through the implementation of the above steps, the synergistic effect of each link in this embodiment forms an accurate analysis process of the crack initiation point from data acquisition, matching time, interpolation calculation to recording results. Each step is not only interrelated, but also lays the foundation for the rigor of experimental data and the reliability of results, greatly promotes the in-depth study of the mechanical behavior of materials, and helps to understand the microstructural changes and mechanical responses of materials under stress. These achievements are also of great value in engineering applications, especially in the design and safety assessment of materials.

[0123] Furthermore, the stage of analyzing the occurrence of cracks in step S34 specifically includes the following steps:

[0124] Step S341: finding the stress value and strain value when the crack occurs by interpolation, marking the crack initiation point on the stress-strain curve, and analyzing the characteristics of the overall stress-strain curve;

[0125] Among them, the overall stress-strain curve is usually divided into several important stages:

[0126] Elastic stage: In this stage, the stress and strain of the material increase proportionally, that is, it follows Hooke's law; the relationship between stress and strain is linear, and the material can return to its original state after the external force is removed;

[0127] Yield stage: When the stress reaches a certain point (yield point), the material begins to undergo plastic deformation, and the stress is no longer proportional to the strain. If cracks occur during this stage, it usually means that the material has undergone significant internal deformation.

[0128] Strengthening stage: After yielding, the material may continue to be subjected to higher stresses and strains due to phenomena such as work hardening, but this may be accompanied by the formation of microcracks;

[0129] Instability stage: In this stage, the material may experience failure, fracture, etc.

[0130] Step S342: Determine whether the crack occurs after the yield point, and check the relationship between the stress value at the crack initiation point and the yield stress:

[0131] If the stress value is less than the yield stress, the crack may be generated in the elastic stage of the material, which means that the microstructural damage of the material begins before yielding.

[0132] If the stress value ≈ yield stress: cracks may be generated near the yield point, and the plastic deformation stage of the material begins, indicating that microcracks are formed after reaching the yield point;

[0133] If the stress value > yield stress: the crack occurs in the nonlinear behavior stage of the material, indicating that the material has undergone significant deformation, and the formation of cracks may be accompanied by plastic flow and brittle fracture;

[0134] Step S343: According to the location of the crack initiation point, comprehensively analyze the characteristics of the elastic, yield and strengthening stages, record the stress value, strain value and corresponding experimental conditions when the crack occurs, and organize the analysis results into an experimental report, including the material behavior characteristics, crack initiation stress, strain value, and clarify the stage of crack generation and its cause.

[0135] Preferably, step S341 of this embodiment performs interpolation and stress-strain curve feature analysis to accurately locate the crack initiation point: the stress and strain values ​​when the crack occurs are obtained by interpolation calculation, and the point is marked on the stress-strain curve to ensure its visibility; the overall stress-strain curve is divided into elastic, yield, strengthening and instability stages to make the analysis more structured and facilitate identification of the mechanical behavior of the material; data visualization: by marking the crack initiation point, the material state when the crack is formed can be intuitively understood, so that subsequent analysis can be carried out more effectively; the performance of the material under different loading stages can be clearly distinguished, and the understanding of the mechanical properties of the material can be expanded. Step S342 determines whether the crack occurs after the yield point. By comparing the stress value at the crack initiation point with the yield stress, the stage characteristics of the crack generation are analyzed to determine whether the crack occurs in the elastic stage, the yield stage or the nonlinear behavior stage; the specific stage of the crack occurrence is clarified, and a quantitative basis is provided for the response of the material under different loading conditions. Predicting material performance: by determining the crack generation state, the potential failure mode of the material during the stress process can be inferred, so as to make improvement suggestions for the design and application of the material; improving material safety: understanding the crack generation stage helps engineers optimize material use strategies to prevent accidental failures caused by cracks in actual applications and enhance the safety of the structure. Step S343 comprehensively analyzes and records the experimental records, organizes the stress value, strain value and experimental conditions when the crack is initiating, and provides accurate information for subsequent analysis or verification. This process helps to form a comprehensive experimental record to ensure that the data is used reasonably; analyze the characteristics of the elastic, yield and strengthening stages on the overall stress-strain curve, and through comparison and summary, form a comprehensive understanding of the material properties; form an experimental report: record the results in detail and summarize the stage characteristics, which helps to generate high-quality experimental reports. The analysis results provide rich data support for subsequent research and promote progress in the field of materials science, especially in crack growth and its control.

[0136] In summary, through the implementation of these steps in this embodiment, the entire analysis process can clearly and accurately identify and analyze the specific stages of crack generation, and reveal the mechanical behavior characteristics of materials under different stress conditions. Each step not only plays an important role in technology, but also provides a scientific basis and guidance for material performance evaluation and engineering applications, which will directly affect material design, optimization and accident prevention, and has far-reaching practical significance.

[0137] like Figure 6 As shown, this embodiment also provides an embodiment of a system for early warning of apparent crack initiation of rock sample under pressure. In this embodiment, the system for early warning of apparent crack initiation of rock sample under pressure is applied to the method for early warning of apparent crack initiation of rock sample under pressure in the above embodiment. The system for early warning of apparent crack initiation of rock sample under pressure includes a model building module 1, an image recognition module 2, and a crack determination module 3 which are electrically connected in sequence.

[0138] Among them, the model construction module 1 is used to obtain historical rock sample surface crack-free images and cracked images to construct a data set, divide the data set into a training set and a test set, use the training set to train the convolutional neural network classification model, and construct an apparent crack recognition model; the image recognition module 2 is used to use high-definition macro photography to collect all frames of each video of the rock sample microscopic surface image, input each frame of the image into the apparent crack recognition model, identify the apparent crack and generate an extended image; the crack determination module 3 is used to determine a point on the stress-strain curve according to the time corresponding to the first apparent crack image, which is the apparent crack initiation point, and determine the stage of appearance of the apparent crack of the rock sample under pressure and the corresponding initiation stress.

[0139] Preferably, the model building module 1 of this embodiment is responsible for collecting and marking the crack-free and cracked images of the rock sample, and constructing an effective data set for training; using the convolutional neural network (CNN) to train the training set, automatically extract crack features, and construct an efficient apparent crack recognition model; dividing the data set into a training set and a test set to ensure the generalization ability of the model during the training process and verifying the model effect; improving the recognition accuracy: through a large amount of training data, the accuracy and robustness of the crack recognition model are improved, thereby providing reliable support for subsequent image recognition and crack detection; laying the foundation for subsequent modules: the constructed apparent crack recognition model is the core of the entire system, ensuring that the subsequent image recognition module can effectively recognize crack images and improve the practicality of the entire system. The image recognition module 2 uses high-definition macro photography technology to collect images of the microscopic surface of the rock sample, providing high-quality raw data for subsequent crack recognition; inputting each frame of the image into the trained apparent crack recognition model, automatically identifying the apparent crack, thereby obtaining real-time recognition results and improving the detection speed; being able to process all frames in the video, ensuring that the crack state of the rock sample can be accurately captured at different time nodes. Real-time monitoring and feedback: Improve the real-time performance of the system so that crack detection can provide instant feedback, timely discover potential structural problems, and reduce accident risks; Enhance data acquisition capabilities: By processing each frame of the image, a large amount of crack data is accumulated to provide data support for subsequent analysis and prediction, thereby improving the effectiveness of the early warning system. Crack determination module 3 accurately determines a point on the stress-strain curve, i.e., the crack initiation point, according to the time corresponding to the first identified apparent crack image; analyzes the stage of rock sample crack occurrence and the corresponding crack initiation stress based on the stress-strain data at the crack initiation point, providing a basis for the performance evaluation of the rock sample; presents the analysis results in a visual way so that decision makers and engineers can quickly obtain important information. Effective risk management: It can accurately determine the stage of crack formation and the corresponding crack initiation stress, providing data support for the management and evaluation of the strength and stability of rock samples; by clarifying the crack initiation point and its related parameters, it helps engineering and technical personnel make more scientific design and maintenance decisions to ensure the safety and stability of the structure.

[0140] In summary, in this embodiment, through the collaborative work of the model building module 1, the image recognition module 2 and the crack determination module 3, the rock sample compression apparent crack initiation warning system can achieve efficient and accurate crack monitoring and warning. The close connection between each module ensures the efficiency and practicality of the entire system, provides an effective monitoring method for geotechnical engineering and materials science research, and has important theoretical research and practical application value.

[0141] Furthermore, if Figure 7 and Figure 8As shown, the rock sample compression apparent crack initiation warning system of this embodiment also includes: a rock sample 4, a camera fixing bracket 5, a camera 6, a loading device 7, and a macro lens 8;

[0142] The rock sample 4 is placed in the middle of the loading device 7, and a plurality of cameras 6 are fixed in the horizontal direction of the rock sample 4. The cameras 6 are fixed by a plurality of camera fixing brackets 5, and a macro lens 8 is installed on the camera 6; the camera 6 records the video of the whole process of the rock sample being compressed, and the macro lenses 8 are evenly arranged in the horizontal direction of the cylindrical rock sample 4 and equidistantly arranged in the vertical direction;

[0143] Preferably, in the operation process of this embodiment, ensure that the experimental table and the loading device 7 are in a horizontal state to ensure the stable placement of the rock sample 4; calibrate the alignment of the camera 6, the macro lens 8 and the loading device 7 to ensure that the viewing angle of the camera 6 can cover the entire surface of the rock sample 4; place the rock sample to be tested (rock sample 4) in the middle position of the loading device 7 to ensure that the bottom of the rock sample 4 is in good contact with the loading device 7; check whether the rock sample 4 is placed horizontally to ensure uniform force. If necessary, make fine adjustments; use multiple camera fixing brackets 5 to fix multiple cameras 6 in the horizontal direction of the rock sample 4; the cameras 6 should be evenly distributed around the rock sample 4 to ensure that the changes on its surface can be observed at different angles; install a macro lens 8 on each camera 6 to enhance the ability to capture fine cracks; adjust the shooting angle and focal length of each camera 6 as needed to ensure that the details of the surface of the rock sample 4 can be clearly recorded; start the recording software on each camera 6 and configure it, including resolution, frame rate, etc., to ensure that high-quality video is recorded; before starting to apply pressure, confirm that all cameras 6 are in normal working condition and start recording the video of the pressure process of the rock sample 4; start the loading device 7, Apply compressive force to the rock sample 4 gradually, and set a fixed loading rate as needed; in the process of applying pressure, observe the video captured by the camera 6 and pay close attention to whether cracks appear on the surface of the rock sample 4; record the video of the entire compression process through the camera 6 to capture the cracks and deformation on the surface of the rock sample 4; save the recorded video data for subsequent analysis; analyze the collected video data frame by frame to capture the occurrence time, location and development process of the cracks in the rock sample 4; after confirming that the experimental goal has been achieved, stop the loading device 7 and remove the rock sample 4 from the loading device 7; turn off the camera 6 and related equipment, and clean and organize the equipment for the next experiment (the principle is shown in the attached figure). Fig. 9 -Attached Fig.10 ).

[0144] In summary, the rock sample compressive apparent crack initiation early warning system of the present embodiment realizes real-time monitoring of crack generation in the rock sample 4 under compressive state through system preparation, fixation of the rock sample 4, installation and setting of the camera 6, video recording and monitoring of the loading process; the precise implementation of each step is crucial for obtaining reliable experimental data and conducting subsequent analysis; the data acquisition system includes a camera fixing bracket 5, a camera 6, and a macro lens 8, which uses high-definition macro photography to collect the microscopic surface image of the rock sample, and uses an apparent crack recognition model based on a convolutional neural network to identify the apparent crack image. According to the time corresponding to the first apparent crack image, a point on the stress-strain curve can be determined, which is the apparent crack initiation point, thereby determining the stage of the apparent crack initiation of the rock sample under compressive state and the corresponding initiation stress.

[0145] In summary, this embodiment can not only accurately identify the initial apparent crack initiation point appearance stage and crack initiation stress, but also identify the subsequent apparent crack initiation point appearance stage and crack initiation stress; this embodiment records the apparent crack propagation process and time, and can determine the interval corresponding to the stress-strain curve.

[0146] like Fig.11 As shown, this embodiment provides an embodiment of an electronic device. In this embodiment, the electronic device 9 includes a processor 91 and a memory 92 coupled to the processor 91.

[0147] The memory 92 stores program instructions for implementing the method for early warning of apparent crack initiation in rock samples under pressure according to any of the above embodiments.

[0148] The processor 91 is used to execute the program instructions stored in the memory 92 to provide early warning of the initiation of apparent cracks in rock samples under pressure.

[0149] The processor 91 may also be referred to as a CPU (Central Processing Unit). The processor 91 may be an integrated circuit chip having the ability to process signals. The processor 91 may also be a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.

[0150] Further, Fig.12The schematic diagram of the structure of the storage medium of an embodiment of the present application is that the storage medium 10 of the embodiment of the present application stores program instructions 101 that can implement all the above methods, wherein the program instructions 101 can be stored in the above storage medium in the form of a software product, including several instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) or a processor to execute all or part of the steps of the methods described in each embodiment of the present application. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, or terminal devices such as a computer, a server, a mobile phone, and a tablet.

[0151] In the several embodiments provided by the present invention, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic, for example, the division of units is only a logical function division, and there may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interfaces, devices or units, which can be electrical, mechanical or other forms.

[0152] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above integrated unit may be implemented in the form of hardware or in the form of software functional units. The above is only an implementation mode of the present invention, and does not limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made by using the contents of the present invention specification and drawings, or directly or indirectly used in other related technical fields, is also included in the patent protection scope of the present invention.

[0153] The specific implementation methods of the invention are described in detail above, but they are only examples, and the invention is not limited to the specific implementation methods described above. For those skilled in the art, any equivalent modification or substitution of the invention is also within the scope of the invention, therefore, the equalization, modification, improvement, etc. made without departing from the spirit and principle of the invention should be included in the scope of the invention.

Claims

1. A method for early warning of apparent crack initiation in rock samples under pressure, characterized in that: The rock sample compression apparent crack initiation early warning method comprises: Obtain historical rock sample surface images with and without cracks to construct a data set, divide the data set into a training set and a test set, use the training set to train a convolutional neural network classification model, and construct an apparent crack recognition model; All frames of each video of the rock sample microscopic surface image are collected using high-definition macro photography, and each frame of the image is input into the apparent crack recognition model to identify the apparent crack and generate an extended image; According to the time corresponding to the first apparent crack image, a point on the stress-strain curve is determined, which is the apparent crack initiation point, and the stage of apparent crack appearance and the corresponding crack initiation stress of the rock sample under pressure are determined.

2. The method for early warning of apparent crack initiation in rock samples under pressure according to claim 1, characterized in that: Construct an apparent crack identification model, including: From the digital image database, we collected labeled images of rock samples with and without cracks on their surfaces, and divided the images into training sets and test sets according to the ratio of 8:

2. Create convolutional layers, activation layers, pooling layers, fully connected layers, and output layers in the convolutional neural network. Create a flattening layer in the fully connected layer to convert the three-dimensional output into a one-dimensional vector. Select the cross entropy loss function, optimizer SGD and evaluation indicators, compile the apparent crack recognition model; use the image data in the training set for training, adjust the model parameters of the apparent crack recognition model, use the test set to evaluate the trained apparent crack recognition model, calculate the accuracy, recall rate and F1 score evaluation indicators of the apparent crack recognition model, and decide whether to fine-tune or optimize the apparent crack recognition model based on the evaluation results.

3. The method for early warning of apparent crack initiation in rock samples under pressure according to claim 2, characterized in that: Compile the apparent crack identification model, including: Import the TensorFlow or Pytorch deep learning library, use the sequential model to build the entire convolutional neural network model structure, and list the specific parameters of each layer; Use the cross entropy loss function to create comparison examples; use stochastic gradient descent to add momentum to accelerate convergence and avoid local optimal solutions. Start with SGD and set a small learning rate. After obtaining preliminary results, adjust according to the performance of the training process; select evaluation indicators such as accuracy, recall, precision, and F1 score; Combine the model with the selected loss function, optimizer, and evaluation metric to prepare it for training; when compiling, make sure to save all relevant settings of the model; use the compiled model to train by calling the training function to pass in the training data, set the hyperparameters, and verify the model performance after each round of training.

4. The method for early warning of apparent crack initiation in rock samples under pressure according to claim 1, characterized in that: High-definition macro photography is used to collect all frames of each video of the rock sample microscopic surface images, including: Ensure that the test bench and loading device are in a horizontal state, and calibrate the alignment of the camera, macro lens and loading device; place the rock sample to be tested in the middle position of the loading device; check whether the rock sample is placed horizontally; Use multiple camera fixing brackets to fix multiple cameras in the horizontal direction of the rock sample; install a macro lens on each camera, and adjust the shooting angle and focal length of each camera; start the recording software on each camera and configure it, including resolution and frame rate; Before starting to apply pressure, confirm that all cameras are in normal working condition and start recording the video of the rock sample under pressure; start the loading device and gradually apply compression force to the rock sample. Set a fixed loading rate as needed; during the process of applying pressure, observe the video captured by the camera and pay close attention to whether cracks appear on the surface of the rock sample; record the video of the entire compression process through the camera to capture the cracks and deformation on the surface of the rock sample; save the recorded video data; The collected video data is analyzed frame by frame to capture the occurrence time, location and development process of the rock sample cracks; after confirming that the experimental objectives have been achieved, the loading device is stopped and the rock sample is removed from the loading device; the camera and related equipment are turned off, and the equipment is cleaned and organized for the next experiment.

5. The method for early warning of apparent crack initiation in rock samples under pressure according to claim 4, characterized in that: Apply compressive forces to the rock sample in a gradual manner, including: Connect and turn on the power of the loading device and accept the control command; the loading device begins to gradually apply compression force to the rock sample according to the preset rate; monitor the display screen of the loading device to observe the change of applied pressure, displacement, and force increase; The loading rate was set to a constant 0.5 MPa / s, and while the pressure was applied, the applied force and time information were recorded; When the set conditions are reached, stop the loading device and record the process of slowing down the pressure to zero; after confirming that it has ended, remove the applied compression force.

6. The method for early warning of apparent crack initiation in rock samples under pressure according to claim 1, characterized in that: The camera's recording software and loading device operate synchronously, including: Install the data acquisition software in the camera's recording software and configure the acquisition rate and required parameters to ensure that each sample has a timestamp; Design a trigger circuit. When the loading device starts to apply pressure, use the pressure sensor to close the trigger switch. Through control logic programming, use TTL signal and set it to high level as the trigger timing. When the activation signal is sent, a signal is sent to the camera's recording start port. The camera supports external triggering via LANC, USB or wireless connection. Start the recording software and confirm to start recording, and the force of the loading device will be automatically applied to the rock sample, triggering the device to start camera recording.

7. The method for early warning of apparent crack initiation in rock samples under pressure according to claim 1, characterized in that: Determine the initiation point of the apparent crack, including: From the first image where the apparent crack is identified, the timestamp of the frame image is recorded. The timestamp is generated during video recording and is synchronized with the data acquisition system. Use the stress-strain data recorded by the data acquisition system, including stress and strain values ​​that change with time; according to the recorded timestamp, find the stress and strain values ​​corresponding to that time in the stress-strain curve data; through the interpolation method, if the crack time observed in the video does not completely match the time recorded in the data, interpolate the data near the time to find the stress and strain values ​​at that moment; Use the data visualization tool MATLAB to draw the stress-strain curve using the updated stress and strain data; mark the crack initiation point found on the stress-strain curve, which is the stress value and corresponding strain value when the crack appears; According to the relationship between the position of this point and the overall stress-strain curve, the stage at which the crack occurs is analyzed; whether it is the elastic stage, yield stage or other obvious material response characteristics; at the same time, whether this point occurs after the yield point of the material, the behavior of the material is judged; the crack initiation stress and experimental conditions are recorded, and the mapping and stage characteristics required in the experimental report are generated.

8. The method for early warning of apparent crack initiation in rock samples under pressure according to claim 7, characterized in that: By interpolating data near time, including: obtaining a timestamp of the recording, the timestamp corresponding to the first identified apparent crack image; extracting time-varying stress and strain data from the data acquisition system, organized in a time series format, recording the stress and strain applied at different time points; Check the timeline recorded by the data acquisition system to find a time point close to the crack timestamp; if a directly matching timestamp does not exist in the acquired data, it is necessary to determine two time points within a small range; Perform interpolation calculations; Record the interpolation results and the stress σ obtained by interpolation c and strain ∈ c , the two values ​​are considered as the stress and strain that should exist when the crack occurs.

9. A system for early warning of apparent crack initiation in rock samples under pressure, which is applied to the method for early warning of apparent crack initiation in rock samples under pressure as claimed in any one of claims 1 to 8, characterized in that: The rock sample compression apparent crack initiation early warning system comprises: The model building module is used to obtain the images of historical rock samples with and without cracks on the surface to build a data set, divide the data set into a training set and a test set, use the training set to train the convolutional neural network classification model, and build an apparent crack recognition model; An image recognition module is used to collect all frames of each video of the microscopic surface image of the rock sample using high-definition macro photography, input each frame of the image into the apparent crack recognition model, identify the apparent crack and generate an extended image; The crack determination module is used to determine a point on the stress-strain curve according to the time corresponding to the first apparent crack image, which is the apparent crack initiation point, and to determine the stage of the apparent crack appearance and the corresponding crack initiation stress of the rock sample under pressure.

10. The rock sample compression apparent crack initiation early warning system according to claim 9, characterized in that: Also includes: Rock sample, camera fixing bracket, camera, loading device, macro lens; Among them, the rock sample is placed in the middle of the loading device, and multiple cameras are fixed in the horizontal direction of the rock sample. The cameras are fixed by multiple camera fixing brackets, and the macro lenses are installed on the cameras; the cameras record the video of the entire process of the rock sample being subjected to pressure, and the macro lenses are evenly arranged along the horizontal direction of the cylindrical rock sample and equidistantly arranged along the vertical direction.

Images