Stress-strain curve finite element numerical simulation method, device and electronic equipment

Abstract

This invention provides a finite element numerical simulation method, apparatus, and electronic device for stress-strain curves, relating to the field of numerical simulation technology. The method includes: obtaining a stress-strain curve through a uniaxial compression test on a specimen; determining the compaction stage of the specimen based on the stress-strain curve; performing a uniaxial compression simulation on a preset initial three-dimensional model, dynamically updating the tangent modulus of the compaction stage of the initial three-dimensional model to obtain a three-dimensional numerical model; and performing finite element numerical simulation of the stress-strain curve based on the three-dimensional numerical model, which can improve the accuracy of the finite element numerical simulation results.

Description

Technical Field

[0001] This invention relates to the field of numerical simulation technology, and in particular to a finite element numerical simulation method, apparatus, and electronic device for stress-strain curves. Background Technology

[0002] The study of rock mechanical behavior is fundamental to underground engineering design and safety assessment. The full stress-strain curves of rock under uniaxial or triaxial compression conditions comprehensively record the entire process of rock from compaction, elastic deformation, crack propagation, to failure, providing crucial information for revealing rock mechanical properties. The compaction stage, as the initial stage of loading, reflects the gradual closure process of primary micropores and microfractures within the rock under pressure. Accurate simulation of the compaction stage is of significant engineering importance for predicting the initial deformation of engineering rock masses and assessing surrounding rock stability.

[0003] However, existing finite element simulation methods generally employ constitutive models with a constant elastic modulus when simulating the full stress-strain curve of rock, assuming that the rock is a linear elastic material before yielding. This approach results in a linearly increasing stress-strain curve from the origin, failing to accurately reflect the nonlinear deformation characteristics of rock stiffness gradually increasing with stress, leading to insufficient accuracy in the finite element numerical simulation results. Summary of the Invention

[0004] To address the problems existing in the prior art, the present invention provides a finite element numerical simulation method, apparatus, and electronic device for stress-strain curves.

[0005] This invention provides a finite element numerical simulation method for stress-strain curves, comprising: Stress-strain curves were obtained by performing uniaxial compression tests on the specimens. The compaction stage of the specimen is determined based on the stress-strain curve. A uniaxial compression simulation is performed on the preset initial three-dimensional model, and the tangent modulus of the initial three-dimensional model in the compaction stage is dynamically updated to obtain a three-dimensional numerical model. Based on the aforementioned three-dimensional numerical model, a finite element numerical simulation of the stress-strain curve was performed.

[0006] According to the finite element numerical simulation method for stress-strain curves provided by the present invention, determining the compaction stage of the specimen based on the stress-strain curve includes: The stress-strain curve is input into the compaction stage identification model to obtain the final strain of the compaction stage of the sample output by the compaction stage identification model; wherein, the compaction stage identification model is trained based on the sample stress-strain curve and the corresponding real label; The compaction stage of the specimen is determined based on the strain at the end of the compaction stage.

[0007] According to the stress-strain curve finite element numerical simulation method provided by the present invention, the compaction stage identification model includes: An input layer is used to receive the stress-strain curve; wherein the stress in the stress-strain curve is stored in a first channel, and the strain in the stress-strain curve is stored in a second channel. A feature extraction layer is used to extract the feature vector of the stress-strain curve based on the first channel and the second channel; A fully connected layer is used to output the final strain of the compaction stage of the specimen based on the eigenvector.

[0008] According to the finite element numerical simulation method for stress-strain curves provided by the present invention, the feature extraction layer includes: The first convolutional branch is used to capture the curvature features of the stress-strain curve based on the first channel and the second channel; The second convolution branch is used to capture the morphological features of the stress-strain curve based on the first channel and the second channel; The third convolution branch is used to capture the trend characteristics of the stress-strain curve based on the first and second channels; A bidirectional long short-term memory network sublayer is used to extract the feature vector of the stress-strain curve based on the multi-scale fusion feature obtained by splicing the curvature feature, the morphological feature and the trend feature.

[0009] According to the stress-strain curve finite element numerical simulation method provided by the present invention, the fully connected layer includes: The first sub-layer is used to map the feature vector into 128-dimensional features and perform ReLU activation and Dropout regularization; The second sublayer is used to output the final strain of the compaction stage of the specimen by linear activation based on the output of the first sublayer.

[0010] According to the finite element numerical simulation method for stress-strain curves provided by the present invention, the method further includes: Based on the evaluation index of the end strain of the compaction stage, the initial learning rate of the compaction stage identification model is determined to be 0.001, the batch size is 64, and training is stopped when the loss on the validation set no longer decreases after 15 consecutive training rounds. The evaluation metrics include mean absolute error, coefficient of determination, and relative error distribution; the learning rate is multiplied by 0.5 every 20 training epochs.

[0011] According to the finite element numerical simulation method for stress-strain curves provided by the present invention, the step of performing uniaxial compression simulation on a preset initial three-dimensional model includes: Fix the lower face of the preset initial 3D model, and apply a 5×10⁻⁶ calculation step to the upper face of the preset initial 3D model. -8 The speed of meters.

[0012] The present invention also provides a finite element numerical simulation device for stress-strain curves, comprising: The curve acquisition module is used to obtain stress-strain curves based on uniaxial compression tests of specimens. The compaction stage determination module is used to determine the compaction stage of the sample based on the stress-strain curve. The model acquisition module is used to perform uniaxial compression simulation on the preset initial three-dimensional model, dynamically update the tangent modulus of the initial three-dimensional model in the compaction stage, and obtain a three-dimensional numerical model. The numerical simulation module is used to perform finite element numerical simulation of stress-strain curves based on the three-dimensional numerical model.

[0013] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the stress-strain curve finite element numerical simulation method as described above.

[0014] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the finite element numerical simulation method for stress-strain curves as described above.

[0015] This invention provides a finite element numerical simulation method, apparatus, and electronic device for stress-strain curves. The method involves obtaining a stress-strain curve through a uniaxial compression test on a specimen, determining the compaction stage of the specimen based on the stress-strain curve, performing a uniaxial compression simulation on a preset initial three-dimensional model, dynamically updating the tangent modulus of the compaction stage of the initial three-dimensional model to obtain a three-dimensional numerical model, and performing finite element numerical simulation of the stress-strain curve based on the three-dimensional numerical model. This method can realistically reproduce the nonlinear compaction deformation characteristics of the specimen, including the progressively closing stage, where the stiffness gradually increases with stress during the progressively closing stage, thus improving the accuracy of the finite element numerical simulation results. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0017] Figure 1This is a flowchart illustrating the finite element numerical simulation method for stress-strain curves provided by this invention.

[0018] Figure 2 This is one of the schematic diagrams illustrating an example of the finite element numerical simulation method for stress-strain curves provided by this invention.

[0019] Figure 3 This is the second example of the finite element numerical simulation method for stress-strain curves provided by this invention.

[0020] Figure 4 This is a schematic diagram of the structure of the stress-strain curve finite element numerical simulation device provided by the present invention.

[0021] Figure 5 This is a schematic diagram of the physical structure of the electronic device provided by the present invention. Detailed Implementation

[0022] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0023] In geotechnical and mining engineering, the study of rock mechanical behavior is fundamental to underground engineering design and safety assessment. The full stress-strain curves of rock under uniaxial or triaxial compression conditions comprehensively record the entire process of rock from compaction, elastic deformation, crack propagation, to failure. The compaction stage, also known as the crack closure stage, is the initial stage of loading and reflects the gradual closure process of the original micropores and microcracks within the rock under pressure. Accurately simulating the compaction stage is of great significance for predicting the initial deformation of engineering rock masses and assessing the stability of surrounding rock.

[0024] The three-dimensional finite element method is the main method for studying the mechanical behavior of continuous media. For example, FLAC3D, a three-dimensional continuous media mechanical numerical analysis software based on the explicit finite difference method, can simulate the stress-strain distribution, plastic zone development, and large deformation of rock masses by solving the motion equations of three-dimensional mesh elements under the action of forces.

[0025] However, existing FLAC3D constitutive models, such as isotropic elastic models and Mohr-Coulomb models, typically assume that rock is a linear elastic material before yielding, i.e., using constant elastic and shear moduli. This results in a linear rise in the stress-strain curve from the origin in uniaxial or triaxial compression simulations, missing the nonlinear compaction stage of actual rock during the initial loading phase due to the gradual closure of micropores and microcracks. In other words, traditional models cannot automatically simulate the strengthening phenomenon of rock stiffness gradually increasing with stress, or the dynamic stiffness growth process of rock stiffness evolving with stress or strain states.

[0026] Accurate identification of the compaction stage endpoint is fundamental to improving the simulation accuracy of the compaction stage. Currently, the endpoint is mainly identified through methods such as observation, volumetric strain analysis, and tangent analysis. Specifically: The observation method involves manually visually identifying the points where the slope of the stress-strain curve changes significantly, thus identifying the end point of the compaction stage. However, this method is highly dependent on the operator's subjective experience, and different people may have significantly different interpretations of the same curve, resulting in poor repeatability. The volumetric strain method identifies the end point of the compaction stage by using the inflection point of the volumetric strain axial stress curve from compression to linear contraction. This method has a clear physical meaning, but it requires accurate measurement of lateral strain, and the inflection point is difficult to locate accurately when there is noise in the data. The tangent method involves fitting a straight line to the linear elastic segment and taking the point where it separates from the compaction stage curve as the inflection point to identify the end point of the compaction stage. However, this method requires manual determination of the range of the linear elastic segment, which involves subjectivity and arbitrariness, and it is difficult to accurately define the curve where the compaction stage and the linear elastic segment transition smoothly.

[0027] The aforementioned traditional methods generally suffer from problems such as strong subjectivity, low efficiency, and non-unique results. Furthermore, errors in manual interpretation can be directly transmitted to subsequent simulations, leading to systematic deviations in the simulation accuracy during the compaction stage.

[0028] Based on this, the following combination Figures 1 to 5 This invention describes the finite element numerical simulation method, apparatus, and electronic device for stress-strain curves.

[0029] Figure 1 This is a flowchart illustrating the finite element numerical simulation method for stress-strain curves provided by this invention, as shown below. Figure 1 As shown, the method includes the following steps.

[0030] Step 101: Obtain the stress-strain curve by conducting a uniaxial compression test on the specimen.

[0031] Here, a specimen refers to a sample used for mechanical testing. The interior of a specimen contains native micropores and microcracks, so the deformation of the specimen under pressure includes a progressive closing stage.

[0032] It should be noted that there are many ways to obtain stress-strain curves based on uniaxial compression tests of specimens, such as loading with a constant displacement rate using an electro-hydraulic servo pressure testing machine, or loading using a stress-controlled method, etc. This embodiment does not limit the specific methods.

[0033] For example, rock blocks taken from the site can be processed into uniaxial compressive strength standard specimens: standard cylinders with a height × base diameter = 100mm × 50mm; the uniaxial compressive strength of the uniaxial compressive strength standard specimens can be tested, and the axial stress-strain curve of the rock under uniaxial compression can be plotted.

[0034] Step 102: Determine the compaction stage of the specimen based on the stress-strain curve.

[0035] The compaction stage refers to the region in the deformation stage of the specimen where the original micropores and microcracks gradually close. The compaction stage is characterized by a nonlinear, upward-concave shape in the stress-strain curve.

[0036] It should be noted that there are many ways to determine the compaction stage of a sample based on the stress-strain curve, such as visual interpretation or analysis of the inflection point of the volumetric strain curve. This embodiment does not limit the methods used.

[0037] Step 103: Perform uniaxial compression simulation on the preset initial three-dimensional model, dynamically update the tangent modulus of the initial three-dimensional model in the compaction stage, and obtain a three-dimensional numerical model.

[0038] The initial three-dimensional model refers to the finite element mesh model established based on the geometric dimensions of the specimen; the three-dimensional numerical model refers to the computational model that can be used to simulate mechanical behavior after the finite element mesh model is dynamically updated by the microscopic elastic modulus.

[0039] It should be noted that, in the uniaxial compression simulation of the preset initial three-dimensional model, the tangent modulus of the initial three-dimensional model is dynamically updated during the compaction stage of the deformation stage.

[0040] A three-dimensional mesh consistent with the sample size can be created using finite element software such as FLAC3D, and boundary conditions can be set to construct a preset initial three-dimensional model.

[0041] There are many ways to apply uniaxial compression to dynamically update the tangent modulus of the initial 3D model during the compaction stage, such as through user-defined or built-in variable update functions. This embodiment does not limit this approach.

[0042] Step 104: Perform finite element numerical simulation of stress-strain curves based on the three-dimensional numerical model.

[0043] Taking rock samples as an example, a three-dimensional numerical model can be used to study the mechanical behavior of rocks of the same type as the samples under other loading conditions without having to repeatedly obtain experimental curves.

[0044] The stress-strain curve finite element numerical simulation method provided in this invention involves obtaining a stress-strain curve based on a uniaxial compression test of a specimen, determining the compaction stage of the specimen based on the stress-strain curve, performing a uniaxial compression simulation on a preset initial three-dimensional model, dynamically updating the tangent modulus of the compaction stage of the initial three-dimensional model to obtain a three-dimensional numerical model, and performing a stress-strain curve finite element numerical simulation based on the three-dimensional numerical model. This method can realistically reproduce the nonlinear compaction deformation characteristics of the specimen, including the progressive closure stage of cracks, in the progressive closure stage of cracks, where the stiffness gradually increases with increasing stress, thus improving the accuracy of the finite element numerical simulation results.

[0045] Based on the above embodiments, determining the compaction stage of the specimen according to the stress-strain curve includes: The stress-strain curve is input into the compaction stage identification model to obtain the final strain of the compaction stage of the sample output by the compaction stage identification model; wherein, the compaction stage identification model is trained based on the sample stress-strain curve and the corresponding real label; The compaction stage of the specimen is determined based on the strain at the end of the compaction stage.

[0046] Among them, the end strain of the compaction stage refers to the strain value corresponding to the inflection point where the stress-strain curve changes from the nonlinear compaction stage to the linear elastic stage; the true label is obtained by marking the end of the compaction stage of the stress-strain curve of the sample.

[0047] In some embodiments, the true label is obtained by processing the end strain of the compaction stage of the sample stress-strain curve using a quantitative method.

[0048] Understandably, identifying the compaction stage of a sample from the stress-strain curve using a compaction stage identification model can avoid the subjectivity and inconsistency of manual interpretation, and improve the efficiency and repeatability of compaction stage identification.

[0049] It should be noted that the stress-strain curve is input into the trained compaction stage identification model. After forward propagation, the compaction stage identification model outputs the predicted value. Through inverse normalization, the actual end strain of the compaction stage is obtained, which provides the basis for the subsequent dynamic updating of the tangent modulus of the compaction stage of the initial three-dimensional model, realizing the full-process automation from intelligent data processing to numerical simulation.

[0050] For example, inverse normalization can be performed using the following formula: ; in, This represents the strain at the end of the compaction stage. For predicted values, This represents the peak strain.

[0051] For example, a feasible training scheme for the compaction stage recognition model may include: The sample stress-strain curve is input into the initial recognition model to obtain the recognition result output by the initial recognition model. Then, based on the recognition result and the true label, the loss function value is calculated. Finally, based on the loss function value, the model parameters of the initial recognition model are updated. The above input process and calculation process are iteratively executed until the loss function converges or the preset number of iterations is reached, thus obtaining the compaction stage recognition model. The preset number of iterations can be set as needed and is not specifically limited here.

[0052] In some embodiments, based on the evaluation index of the strain at the end of the compaction stage, the initial learning rate of the compaction stage identification model is determined to be 0.001, the batch size is 64, and training is stopped when the loss on the validation set no longer decreases after 15 consecutive training epochs. The evaluation metrics include mean absolute error, coefficient of determination, and relative error distribution; the learning rate is multiplied by 0.5 every 20 training epochs.

[0053] Specifically, the mean absolute error (MAE) refers to the average absolute deviation between the predicted and actual values; the coefficient of determination (R²) is used to measure the model's ability to explain data variation.

[0054] For example, the relative error distribution can be obtained by statistically analyzing the proportion of samples with relative errors within ±5% or ±10%.

[0055] It should be noted that the loss function for regression tasks can be the mean squared error (MSE). ; Where N is the batch sample size. These are the model's predicted values. This is a real label.

[0056] The optimizer can be Adam.

[0057] In some embodiments, the model performance is evaluated on a test set. Test results show that the MAE of the compaction stage identification model provided in this embodiment of the invention is 0.008, and R0 is... 2 The relative error reached 0.98, with over 90% of the samples having a relative error within ±5%, significantly outperforming traditional methods and other baseline models.

[0058] Understandably, the early stopping mechanism, which stops training when the validation set loss no longer decreases after 15 consecutive training epochs, can prevent the model from continuing to train when the validation set loss no longer improves, thus effectively avoiding overfitting.

[0059] Understandably, the learning rate decay strategy of multiplying the learning rate by 0.5 every 20 training epochs enables the model to finely adjust parameters with smaller steps in the later stages of training, which is beneficial for the loss function to converge smoothly to the global optimum.

[0060] Combining the two approaches can improve the model's generalization ability while ensuring sufficient training, thereby obtaining a high-precision compaction stage recognition model.

[0061] Based on any of the above embodiments, the compaction stage identification model includes: An input layer is used to receive the stress-strain curve; wherein the stress in the stress-strain curve is stored in a first channel, and the strain in the stress-strain curve is stored in a second channel. A feature extraction layer is used to extract the feature vector of the stress-strain curve based on the first channel and the second channel; A fully connected layer is used to output the final strain of the compaction stage of the specimen based on the eigenvector.

[0062] The first channel, also known as the stress channel, refers to the tensor dimension that stores the normalized stress sequence; the second channel, also known as the strain channel, refers to the tensor dimension that stores the normalized strain sequence.

[0063] The eigenvectors of a stress-strain curve are high-dimensional numerical representations that can characterize the local and global morphology of the curve.

[0064] Based on any of the above embodiments, the feature extraction layer includes: The first convolutional branch is used to capture the curvature features of the stress-strain curve based on the first channel and the second channel; The second convolution branch is used to capture the morphological features of the stress-strain curve based on the first channel and the second channel; The third convolution branch is used to capture the trend characteristics of the stress-strain curve based on the first and second channels; A bidirectional long short-term memory network sublayer is used to extract the feature vector of the stress-strain curve based on the multi-scale fusion feature obtained by splicing the curvature feature, the morphological feature and the trend feature.

[0065] The first convolution branch can also be called the curvature convolution branch, the fine part convolution branch, etc.; the second convolution branch can also be called the morphological convolution branch, the medium part convolution branch, etc.; and the third convolution branch can also be called the trend convolution branch, the overall convolution branch, etc.

[0066] It should be noted that the first, second, and third convolutional branches are three parallel convolutional branches, each extracting features at different scales. In other words, curvature features, morphological features, and trend features are features at different scales.

[0067] For example, to capture subtle fluctuations and local curvature changes in the curve, the kernel size of the first convolutional branch can be set to 3. To extract medium-range morphological features, such as the average slope of a segment, the kernel size of the second convolutional branch can be set to 7. To perceive the overall trend and general direction of the curve, the kernel size of the third convolutional branch can be set to 15.

[0068] Among them, curvature features refer to features that reflect the degree of local curvature and subtle fluctuations of the curve, morphological features refer to features that reflect the average slope of the curve over a medium range, trend features refer to features that reflect the overall direction and general shape of the curve, and multi-scale fusion features refer to the comprehensive feature representation formed by splicing the feature maps output from the three branches in the channel dimension.

[0069] The medium range can be a line segment composed of several consecutive data points.

[0070] Understandably, the compaction stage identification model extracts features at different scales through parallel first, second, and third convolutional branches, which can simultaneously focus on the local curvature changes and overall morphological features of the stress-strain curve, thereby improving the accuracy of the compaction stage endpoint location.

[0071] In some embodiments, each convolutional branch of the compaction stage identification model includes two convolutional layers, a ReLU activation function, batch normalization, and a max-pooling layer. The max-pooling layer has a pooling kernel size of 2 for dimensionality reduction.

[0072] It should be noted that after obtaining the curvature feature, the morphological feature, and the trend feature, the three can be spliced ​​together in the channel dimension to achieve the fusion of multi-scale features and obtain multi-scale fused features.

[0073] Among them, the Bidirectional Long Short-Term Memory (Bi-LSTM) sublayer can also be called the dependency capture layer, which is used to capture long-distance dependencies between different positions in the stress-strain curve, enhancing the model's understanding of the overall shape and contextual information of the curve.

[0074] Understandably, the hidden state of each time step in the bidirectional long short-term memory network sublayer contains both past and future contextual information, which can capture the long-distance dependencies of the stress-strain curve and enhance the compaction stage identification model's understanding of the overall shape of the stress-strain curve.

[0075] Based on any of the above embodiments, the fully connected layer includes: The first sub-layer is used to map the feature vector into 128-dimensional features and perform ReLU activation and Dropout regularization; The second sublayer is used to output the final strain of the compaction stage of the specimen by linear activation based on the output of the first sublayer.

[0076] It should be noted that the hidden state of the last time step of the bidirectional long short-term memory network sublayer can be input into the fully connected layer, or the hidden states of all time steps of the bidirectional long short-term memory network sublayer can be averaged and then input into the fully connected layer.

[0077] The first sublayer, also known as the hidden layer, is used to perform non-linear transformations of features and prevent overfitting. In some embodiments, the first sublayer includes 128 neurons with Dropout=0.3.

[0078] The second layer, also known as the output layer, is used to output the predicted end strain of the compaction stage of the specimen. In some embodiments, the second sublayer includes one neuron.

[0079] Understandably, the first and second sub-layers can introduce regularization mechanisms while maintaining expressive power, thereby improving the model's generalization performance.

[0080] Specifically, such as Figure 2 As shown, the input layer can be used to receive a 256×2 stress-strain sequence; The first convolutional branch may include ConvID(3,32), BatchNorm+ReLU, ConvID(3,64)BatchNorm+ReLU, MaxPoolID(2). The second convolutional branch may include ConvID(7,32), BatchNorm+ReLU, ConvID(7,64)BatchNorm+ReLU, MaxPoolID(2). The third convolutional branch may include ConvID(15,32), BatchNorm+ReLU, ConvID(15,64)BatchNorm+ReLU, MaxPoolID(2). The input to the feature stitching layer can be (128, 64), and the output can be (128, 192); the feature stitching layer is used to obtain multi-scale fusion features based on curvature features, morphological features, and trend features. The input to the Bi-LSTM layer can be (128, 192), and the output can be (128, 256). The input to the global average pooling layer can be (128, 256), and the output can be (256); the global average pooling layer is used to process the feature vector output by the Bi-LSTM layer. The input of the first sub-layer of the fully connected layer can be (256), and the output can be (128). The input to the output layer can be (128), and the output can be (1).

[0081] Based on any of the above embodiments, the step of performing uniaxial compression simulation on the preset initial three-dimensional model includes: Fix the lower face of the preset initial 3D model, and apply a 5×10⁻⁶ calculation step to the upper face of the preset initial 3D model. -8 The speed of meters.

[0082] For example, a rock specimen with the same dimensions as the experimental test was constructed using the FLAC3D numerical simulation method. The rock specimen adopted the Strain-Softening Mohr-Coulomb Model (SSMC) constitutive model, and the numerical model is as follows: Figure 3 As shown, uniaxial loading can be achieved by fixing the lower end face of the rock sample and applying a speed of 5e-8 m / step to the upper end face.

[0083] The model mesh should be as fine as possible while taking into account computational efficiency. In some embodiments, the maximum side length of the mesh is 0.002m and the minimum side length is 0.001m. This can more accurately capture the stress-strain distribution and local deformation characteristics inside the sample while ensuring computational convergence and reasonable time consumption, thereby improving the accuracy and reliability of finite element numerical simulation.

[0084] In some embodiments, the sample stress-strain curve is obtained based on uniaxial compression tests of specimens with different lithologies.

[0085] It should be noted that uniaxial compression tests were conducted on samples with different lithologies to obtain sample stress-strain curves, which were then used to train a compaction stage identification model. This improved the model's generalization ability to curves of various lithologies and the accuracy of identifying the end point of the compaction stage.

[0086] Based on the above embodiments, the method further includes: The original curves were obtained by uniaxial compression tests on samples with different lithologies. The target stress range of the original curve is resampled at equal intervals to obtain a target data point sequence, and the corresponding sample stress-strain curve is obtained based on the target data point sequence.

[0087] For example, samples of different lithologies may include sandstone samples, granite samples, marble samples, shale samples, and infill samples, etc. The number of samples should not be less than 1,000.

[0088] For example, a uniaxial compression test based on a prepared rock sample can also be called a uniaxial compressive strength test. After the test, the original curve of the rock sample can be plotted. Based on this original curve, the uniaxial compressive strength of the rock sample, also known as the peak strength σucs, is 71.14 MPa, the elastic modulus E is 15.76 GPa, and the peak strain εucs is 0.7406%, etc. The elastic modulus can be taken as the slope of the stress-strain curve within the range of 0.5σucs ± 1 MPa.

[0089] In some embodiments, in order to eliminate random noise during the uniaxial compression test and retain the main morphological features of the original curve, the original curve can be denoised using methods such as Savitzky-Golay filtering or wavelet denoising before resampling the target stress range of the original curve at equal intervals.

[0090] Understandably, resampling the target stress range of the original curve at equal intervals to obtain the target data point sequence can reduce the risk of data deviation and feature distortion caused by the uneven spacing of the original data points, improve the comparability and consistency of stress-strain curves between different samples, and provide a foundation for the standardized processing of subsequent automatic extraction of mechanical parameters and training of machine learning models.

[0091] In some embodiments, resampling the target stress range of the original curve at equal intervals to obtain a target data point sequence includes: The original curve is resampled at equal intervals from zero to the target stress range of 0.6 times the peak stress to obtain a target data point sequence of fixed length N=256.

[0092] In this embodiment, based on the clear fact that the compaction stage occurs between 0 and 0.5 times the peak stress, a curve segment from zero to 0.6 times the peak stress is selected. This can preserve the key information of the compaction stage and part of the elastic segment while avoiding high noise and unstable regions including the post-peak failure segment, thereby improving the robustness and consistency of feature extraction.

[0093] It is understandable that resampling the target stress range at equal intervals and unifying it into a data point sequence of fixed length N=256 is a suitable length of N=256. This can preserve the main morphological features of the curve while controlling the data dimension and computational overhead, thus achieving a balance between accuracy and efficiency.

[0094] Based on any of the above embodiments, obtaining the corresponding sample stress-strain curve based on the target data point sequence includes: Determine the peak stress and peak strain of the original curve; The stress in the target data point sequence is normalized according to the peak stress, and the strain in the target data point sequence is normalized according to the peak strain to obtain a normalized data point sequence. The corresponding sample stress-strain curve is obtained based on the normalized data point sequence.

[0095] The normalized data point sequence is the dimensionless data point sequence corresponding to the target data point sequence. The normalized data point sequence is used to eliminate the influence of dimensions and amplitude caused by differences in strength and deformation capacity of different lithological samples, making the stress-strain curves of different samples comparable under a unified normalized scale.

[0096] For example, the stress in the target data point sequence can be normalized according to the peak stress using the following formula: σ ∗ =σ / σ ucs ; Where, σ ∗ This is the normalized stress, where σ is the stress in the target data point sequence. ucs It is the peak stress.

[0097] For example, the strain in the target data point sequence can be normalized according to the peak strain using the following formula: ε ∗ =ε / ε ucs ; Where, ε ∗ This is the normalized strain, where ε is the strain in the target data point sequence. ucs It is the peak strain.

[0098] It should be noted that the normalized curve starts at (0,0) and terminates at (0.6,σ). ε =0.6).

[0099] Understandably, normalizing the stress in the target data point sequence based on peak stress and normalizing the strain in the target data point sequence based on peak strain results in a normalized data point sequence. This can eliminate the influence of dimensions and amplitude caused by differences in strength and deformation capacity of different lithological samples, aligning the stress-strain curves of all samples on a uniform [0,1] dimensionless scale. This avoids high-stress or high-strain samples numerically dominating the subsequent analysis process, while preserving the relative morphological characteristics of the curves at each normalization stage.

[0100] Based on this, the corresponding sample stress-strain curves are obtained from the normalized data point sequence, and a sample dataset for training the model is constructed. This provides a foundation for subsequent cross-lithology curve morphology clustering and mechanical parameter prediction, and significantly improves the generalization ability and prediction accuracy of the compaction stage identification model trained on the sample dataset for samples of different strength grades.

[0101] Based on any of the above embodiments, the method further includes: Determine the peak stress and peak strain of the original curve, and determine the elastic modulus of the linear elastic segment of the specimen based on the peak stress and peak strain; Based on the sliding interval, linear fitting is sequentially performed on each segment of the original curve to obtain the linear fitting determination coefficient R of the sliding interval. 2 ; If the linear fitting determination coefficient is less than the target threshold, the true label is obtained by marking the end point of the compaction stage of the sample stress-strain curve based on the sliding interval.

[0102] The linear elastic segment refers to the interval in the deformation stage of the specimen that corresponds to the compaction stage, where stress and strain have an approximately linear relationship.

[0103] It should be noted that the elastic modulus of the rock can be obtained based on the peak stress and peak strain, so as to calibrate the elastic modulus of the linear elastic segment of the sample.

[0104] In some embodiments, 0.5σ can be calculated. ucs The slope of the stress-strain curve within the range of ±1 MPa yields the elastic modulus of the linear elastic segment of the specimen.

[0105] Understandably, compared to subjectively selecting the linear elastic segment, this embodiment calculates the standard slope by selecting a linear interval near the peak stress, which can effectively avoid the interference caused by small fluctuations in the elastic modulus within the linear elastic segment during actual testing. This provides a unified benchmark threshold for the end point of the compaction stage of the sample stress-strain curve based on the sliding interval, thereby improving the accuracy and robustness of compaction stage identification and labeling.

[0106] The sliding interval refers to a continuous window of data points on the stress-strain curve, starting from the initial point and gradually increasing in length along the sliding direction. The target threshold is the critical value of the coefficient of determination used to determine whether the data points within the sliding interval have a good linear relationship. When the coefficient of determination R for linear fitting of the sliding interval... 2 When the value is below the target threshold, it indicates that the data points within the sliding interval have deviated from the linear relationship, including the critical points in the linear elastic segment and the compaction stage. Specifically, this means that the sliding interval has shifted downwards from the linear elastic segment of the stress-strain curve, for example, from 0.5σ... ucs Initially, when the coefficient of determination R for linear fitting of the sliding interval... 2 When the value is less than the target threshold, it indicates that the data points within the sliding interval have deviated from the linear relationship, including the critical points of the linear elastic segment and the compaction stage.

[0107] In some embodiments, before sequentially performing linear fitting on each segment of the original curve based on the sliding interval, the method further includes: The starting reference point is determined based on the peak stress and the preset coefficient; The sliding interval is determined based on the starting reference point, the preset step size, and the preset interval width.

[0108] The initial reference point is used to determine the starting position of the sliding interval.

[0109] In some embodiments, the starting reference point can be obtained by multiplying the corresponding peak stress by a preset coefficient. For example, the preset coefficient can be 0.4-0.6.

[0110] It should be noted that the preset step size and preset interval width can be set according to the actual working conditions, and this embodiment does not limit them.

[0111] For example, the sliding interval can be 0.5σ ucs Within the range of -y±x MPA, y is sequentially determined to be 0.5, 1, 1.5, 2… based on a preset step size of 0.5, and x is set to 0.5, which is half the width of the preset range. This allows for sequential linear fitting of each segment of the stress-strain curve, yielding the linear fitting determination coefficient R0. 2 If R 2 If the target threshold is less than 0.9, then the midpoint ε of the corresponding interval is taken. a This represents the maximum strain during the compaction stage.

[0112] For example, 0.5σ can be sequentially adjusted using the sliding interval. ucs -0.5±0.5MPA range, 0.5σ ucs -1.0±0.5MPA range, 0.5σ ucs The range of -1.5 ± 0.5 MPa and 0.5σ ucsLinear fitting was performed in the range of -2.0 ± 0.5 MPa, and R0 was obtained. 2 value.

[0113] If 0.5σ ucs R in the range of -2.0±0.5MPa 2 If the strain value is less than 0.9, then the strain value corresponding to the midpoint of the sliding interval is taken as the maximum strain during the compaction stage; if 0.5σ ucs R in the range of -2.0±0.5MPa 2 >0.9 or R 2 =0.9, then continue with 0.5σ ucs -2.5±0.5MPA range, 0.5σ ucs Linear fitting is performed within a sliding interval of -3.0±0.5 MPa until the R value of the sliding interval is satisfied. 2 <0.9. The strain value ε corresponding to the midpoint of the interval that meets the requirements. a That is, the inflection point where the stress-strain curve transitions from the nonlinear compaction stage to the linear elastic stage.

[0114] Where, σ = 9.5 × 10 7 (ε+1.03×10) -4 ) 2.79 +0.66, R 2 =0.9996.

[0115] It is understandable that nonlinear fitting of the compaction stage of the experimental stress-strain curve can introduce dynamic micro-elastic modulus into the simulation to simulate the continuous change of tangential modulus during the compaction stage. However, the stress-strain curves of different rock samples are different, so the above nonlinear fitting process needs to be repeated when simulating the stress-strain curves of different rock samples.

[0116] In some embodiments, the input to the input layer is a preprocessed stress-strain curve sequence with a shape of (256,2), where 256 is the sequence length and the two channels correspond to the normalized stress σ∗ and the normalized strain ε∗, respectively.

[0117] In some embodiments, stress-strain curves are obtained by performing uniaxial compression tests on the specimen to determine the nonlinear relationship between the tangent modulus and strain during the compaction phase of the specimen.

[0118] The compaction stage, also known as the initial loading stage or crack closure stage, refers to the region in the deformation process of a specimen where the original micropores and microcracks gradually close. The compaction stage is characterized by a nonlinear, upward-concave shape in the stress-strain curve.

[0119] It should be noted that the compaction stage is usually completed at a low stress level. Once most of the original microcracks have closed, the stress-strain curve will transition to the linear elastic stage.

[0120] The nonlinear relationship between tangent modulus and strain refers to the functional expression of the dynamic change of tangent modulus with strain. This nonlinear relationship is used to dynamically update the microscopic elastic modulus of the initial three-dimensional model to obtain a three-dimensional numerical model, so that the three-dimensional numerical model can reproduce the stiffness evolution process of rock compaction.

[0121] In some embodiments, a uniaxial compression simulation is performed on a preset initial three-dimensional model to monitor the strain of the initial three-dimensional model.

[0122] The initial three-dimensional model is used to simulate the macroscopic continuous mechanical behavior of the corresponding medium.

[0123] The initial three-dimensional model refers to the finite element model established based on the geometric dimensions of the specimen; the three-dimensional numerical model refers to the computational model that can be used to simulate mechanical behavior after the finite element model is dynamically updated by the microscopic elastic modulus.

[0124] It should be noted that a uniaxial compression simulation is performed on the preset initial three-dimensional model, and the strain of the initial three-dimensional model is monitored during the compaction stage of the deformation stage.

[0125] A simulated specimen with the same dimensions as the original specimen was established using the Strain-Softening Mohr-Coulomb Model (SSMC), and boundary conditions were set to construct a pre-defined initial three-dimensional model.

[0126] There are many ways to apply uniaxial compression and monitor the strain of the initial 3D model, such as recording axial strain in real time through FISH scripts or calculating strain in a built-in measurement area at regular intervals. This embodiment does not limit this method.

[0127] In some embodiments, the microelastic modulus of the initial three-dimensional model is dynamically updated based on the strain and the nonlinear relationship to obtain a three-dimensional numerical model.

[0128] It should be noted that the microscopic elastic modulus of the initial 3D model, calculated based on strain and nonlinear relationships, can be assigned to the initial 3D model in real time using the FISH script. In this way, under the dynamic update mechanism, the model can reproduce the mechanical behavior of the gradually increasing tangential modulus during the compaction stage and the constant modulus during the linear elastic stage, thereby obtaining a full stress-strain curve that closely matches the experimental curve, resulting in a 3D numerical model that can be used for subsequent analysis.

[0129] Based on the above embodiments, determining the nonlinear relationship between the tangent modulus and strain during the compaction stage of the specimen includes: Identify the compaction curve segment corresponding to the compaction stage of the specimen in the stress-strain curve; The nonlinear relationship between the tangent modulus and strain during the compaction stage of the specimen is determined based on the compaction curve segment.

[0130] The compaction curve segment can also be referred to as the compaction stage.

[0131] Understandably, first identifying the compaction curve segment corresponding to the compaction stage of the specimen in the stress-strain curve, and then determining the nonlinear relationship between the tangent modulus and strain in the compaction stage of the specimen, can eliminate the interference of linear elastic segment and subsequent stage data on the nonlinear constitutive relationship, and improve the fitting accuracy and physical reality of the nonlinear relationship between the tangent modulus and strain.

[0132] Based on any of the above embodiments, determining the nonlinear relationship between the tangent modulus and strain of the sample during the compaction stage according to the compaction curve segment includes: The stress-strain relationship of the compaction curve segment is obtained by performing nonlinear fitting on the compaction curve segment. Based on the stress-strain relationship, a nonlinear relationship between the tangent modulus and strain during the compaction stage of the specimen is obtained by differential processing.

[0133] It should be noted that fitting tests can be performed on the fitting functions in the pre-established multi-parameter nonlinear regression model library based on the compaction curve segment, based on R... 2 The constitutive equation for the compaction stage is determined from the fitting function; the multi-parameter nonlinear regression model library includes at least two fitting functions established for the differences in compaction characteristics of different lithologies such as sandstone, granite, and mudstone.

[0134] For example, the fitting function may include an exponential function: ; Exponential function: ; Polynomial: .

[0135] Taking the stress-strain curve corresponding to the strain range of 0~0.402% in the compaction curve segment as an example, we can perform fitting and trial calculations on the above different functions, using R... 2 As the primary evaluation metric, the required R-value after fitting is... 2 ≥0.90, preferred R 2 ≥0.95. Select R. 2The highest function serves as the constitutive equation for the final compaction stage.

[0136] The core fitting algorithm employs the non-linear least squares method, specifically using the Levenberg-Marquardt (LM) algorithm or the Gauss-Newton algorithm for iterative solution, to ensure that the sum of squared residuals between the fitted curve and the experimental data is minimized.

[0137] Based on this, the stress-strain relationship of the compaction curve segment of the rock stress-strain curve is obtained by comparing the fitted curve obtained from the nonlinear fitting with the original data: σ = 9.5 × 10 7 (ε+1.03×10) -4 ) 2.79 +0.66.

[0138] In some embodiments, before performing nonlinear fitting processing based on the compaction curve segment, the method further includes: The strain of the compaction curve segment is dimensionless.

[0139] It should be noted that the horizontal axis strain can be restored to a dimensionless value, for example, 0.4% can be restored to 0.004, in order to reduce the risk of it affecting the accuracy of curve fitting.

[0140] In some embodiments, after curve fitting is completed, the stress unit is converted to the basic unit Pa. In this embodiment, 1 MPa = 1 × 10⁻⁶. 6 Pa, therefore, the final fitted stress-strain relationship is: σ = 9.5 × 10 13 (ε+1.03×10) -4 ) 2.79 +6.6×10 5 .

[0141] It should be noted that, in this embodiment, based on the definition of tangent modulus as the slope of the axial stress-axial strain curve, the nonlinear relationship between tangent modulus and strain during the compaction stage of the specimen is obtained by differentiating the stress-strain relationship: E = 2.65 × 10 14 (ε+1.03×10) -4 ) 1.79 .

[0142] The unit of tangent modulus is Pa.

[0143] Understandably, by differentiating the stress-strain relationship to obtain the nonlinear relationship between the tangent modulus and strain during the compaction stage of the specimen, a continuous analytical expression of the tangent modulus with strain can be directly obtained. This provides a basis for real-time dynamic assignment in finite element simulation, avoiding errors caused by numerical difference, while ensuring the accuracy and computational efficiency of modulus updates.

[0144] It should be noted that the rock compaction stage is caused by the closure of the original pores and fissures in the rock. During the rock compaction stage, the rock stress-strain curve is concave upward, that is, the rock tangential modulus continuously increases, and the rock is in the strengthening stage.

[0145] When using the FLAC3D numerical simulation method to reproduce the compaction stage of rocks, the Strain-Softening Mohr-Coulomb Model (SSMC) in FLAC3D can simulate the macroscopic mechanical properties of rocks by setting corresponding mechanical parameters for finite elements. Based on this, the macroscopic mechanical properties of rocks can be reflected by microscopic mechanical parameters. Specifically, the macroscopic mechanical parameters of rocks can be calibrated by setting microscopic mechanical parameters such as the microscopic elastic modulus, element Poisson's ratio, element cohesion, element tensile strength, element internal friction angle, and softening parameter, so as to accurately simulate the real macroscopic and microscopic mechanical properties of rocks.

[0146] To adjust the finite element numerical model, based on any of the above embodiments, the step of dynamically updating the microelastic modulus of the initial three-dimensional model based on the strain and the nonlinear relationship includes: The strain of the initial three-dimensional model is monitored in real time. When the strain falls within the compaction curve segment, the microelastic modulus e of the initial three-dimensional model is dynamically updated according to the following formula: e=λ[2.65×10 14 (ε+1.03×10) -4 ) 1.79 ]; in, It is a correction factor. It is the strain of the initial three-dimensional model obtained through real-time monitoring.

[0147] The microscopic elastic modulus of the initial three-dimensional model may include the microscopic tangent modulus.

[0148] It should be noted that the script can monitor the rock strain value in real time during the simulation process, and calculate the micro elastic modulus value of the initial three-dimensional model corresponding to the compaction stage using the real-time monitored strain value. The micro elastic modulus is updated in real time to achieve accurate simulation of the rock compaction stage.

[0149] For example, scripts can be written in the FISH language of FLAC3D to monitor the rock strain value in real time during the simulation process, and then scripts can be written in the FISH language to introduce the micro-elastic modulus values ​​of different stages into SSMC to achieve accurate simulation of the rock compaction stage.

[0150] With the strain range of the rock compaction stage as 0≤ Taking ≤0.00402 as an example, during simulation calibration: If 0≤ If ≤0.00402, then: e=λ[2.65×10 14 (ε+1.03×10) -4 ) 1.79 ]; like >0.00402, the rock enters the linear elastic stage, and the microelastic modulus remains unchanged, then: e=λ[2.65×10 14 (4×10) -3 +1.03×10 -4 ) 1.79 ].

[0151] To adjust other micromechanical parameters of the finite element numerical model, based on any of the above embodiments, after dynamically updating the microelastic modulus of the initial three-dimensional model based on the strain and the nonlinear relationship, the method further includes: The microscopic parameters of the initial three-dimensional model are calibrated until the difference between the macroscopic mechanical parameters of the model and the macroscopic mechanical parameters is within the target range; The microscopic parameters include at least one of the following: microscopic elastic modulus correction factor, Poisson's ratio, cohesion, tensile strength, and internal friction angle.

[0152] It should be noted that the microscopic parameters of the initial 3D model can be calibrated using a trial-and-error method. By adjusting the microscopic mechanical parameters of the rock in the model, the macroscopic mechanical parameters of the model can be calibrated until the macroscopic mechanical parameters presented by the model match the experimental results.

[0153] For example, the microscopic elastic modulus correction coefficient λ and Poisson's ratio of the unit are matched to the nonlinear characteristics of the compaction stage, and the macroscopic elastic modulus and Poisson's ratio are matched to the uniaxial compressive strength and peak strain. In this way, the post-peak cohesion and internal friction angle softening parameters are continuously adjusted to match the post-peak mechanical behavior of the rock. This is until the final comparison result of the experimental stress-strain curve and the simulated stress-strain curve is obtained.

[0154] The microscopic parameters of the model used in this simulation study after calibration are shown in the table below.

[0155] The comparison of the mechanical parameters of the simulated and experimental coal and rock samples after calibration is shown in the table below.

[0156] This demonstrates that the simulation method and the experimental results are in good agreement.

[0157] The stress-strain curve finite element numerical simulation device provided by the present invention is described below. The stress-strain curve finite element numerical simulation device described below can be referred to in correspondence with the stress-strain curve finite element numerical simulation method described above.

[0158] Figure 4 This is a schematic diagram of the structure of the stress-strain curve finite element numerical simulation device provided by the present invention, as shown below. Figure 4 As shown, the device includes: Curve acquisition module 410 is used to obtain stress-strain curves based on uniaxial compression tests of specimens. The compaction stage determination module 420 is used to determine the compaction stage of the sample based on the stress-strain curve. The model acquisition module 430 is used to perform uniaxial compression simulation on the preset initial three-dimensional model, dynamically update the tangent modulus of the initial three-dimensional model in the compaction stage, and obtain a three-dimensional numerical model. The numerical simulation module 440 is used to perform finite element numerical simulation of stress-strain curves based on the three-dimensional numerical model.

[0159] Figure 5 An example is a schematic diagram of the physical structure of an electronic device, such as... Figure 5 As shown, the electronic device may include: a processor 510, a communication interface 520, a memory 530, and a communication bus 540, wherein the processor 510, the communication interface 520, and the memory 530 communicate with each other through the communication bus 540. The processor 510 can call logical instructions in the memory 530 to execute a stress-strain curve finite element numerical simulation method. This method includes: obtaining a stress-strain curve by performing a uniaxial compression test on the specimen; determining the compaction stage of the specimen based on the stress-strain curve; performing a uniaxial compression simulation on a preset initial three-dimensional model, dynamically updating the tangent modulus of the compaction stage of the initial three-dimensional model to obtain a three-dimensional numerical model; and performing a stress-strain curve finite element numerical simulation based on the three-dimensional numerical model.

[0160] Furthermore, the logical instructions in the aforementioned memory 530 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0161] On the other hand, the present invention also provides a computer program product, which includes a computer program that can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the stress-strain curve finite element numerical simulation method provided by the above methods. The method includes: obtaining a stress-strain curve by performing a uniaxial compression test on a specimen; determining the compaction stage of the specimen according to the stress-strain curve; performing a uniaxial compression simulation on a preset initial three-dimensional model, dynamically updating the tangent modulus of the compaction stage of the initial three-dimensional model to obtain a three-dimensional numerical model; and performing a stress-strain curve finite element numerical simulation based on the three-dimensional numerical model.

[0162] In another aspect, the present invention also provides a non-transitory computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the stress-strain curve finite element numerical simulation method provided by the above methods. The method includes: obtaining a stress-strain curve by performing a uniaxial compression test on a specimen; determining the compaction stage of the specimen based on the stress-strain curve; performing a uniaxial compression simulation on a preset initial three-dimensional model, dynamically updating the tangent modulus of the compaction stage of the initial three-dimensional model to obtain a three-dimensional numerical model; and performing a stress-strain curve finite element numerical simulation based on the three-dimensional numerical model.

[0163] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0164] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0165] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A stress-strain curve finite element numerical simulation method, characterized by, include: Stress-strain curves were obtained by performing uniaxial compression tests on the specimens. The compaction stage of the specimen is determined based on the stress-strain curve. A uniaxial compression simulation is performed on the preset initial three-dimensional model, and the tangent modulus of the initial three-dimensional model in the compaction stage is dynamically updated to obtain a three-dimensional numerical model. Based on the aforementioned three-dimensional numerical model, a finite element numerical simulation of the stress-strain curve was performed.

2. The finite element numerical simulation method for stress-strain curves according to claim 1, characterized in that, Determining the compaction stage of the specimen based on the stress-strain curve includes: The stress-strain curve is input into the compaction stage identification model to obtain the final strain of the compaction stage of the sample output by the compaction stage identification model; wherein, the compaction stage identification model is trained based on the sample stress-strain curve and the corresponding real label; The compaction stage of the specimen is determined based on the strain at the end of the compaction stage.

3. The finite element numerical simulation method for stress-strain curves according to claim 2, characterized in that, The compression stage identification model includes: An input layer is used to receive the stress-strain curve; wherein the stress in the stress-strain curve is stored in a first channel, and the strain in the stress-strain curve is stored in a second channel. A feature extraction layer is used to extract the feature vector of the stress-strain curve based on the first channel and the second channel; A fully connected layer is used to output the final strain of the compaction stage of the specimen based on the eigenvector.

4. The finite element numerical simulation method for stress-strain curves according to claim 3, characterized in that, The feature extraction layer includes: The first convolutional branch is used to capture the curvature features of the stress-strain curve based on the first channel and the second channel; The second convolution branch is used to capture the morphological features of the stress-strain curve based on the first channel and the second channel; The third convolution branch is used to capture the trend characteristics of the stress-strain curve based on the first and second channels; A bidirectional long short-term memory network sublayer is used to extract the feature vector of the stress-strain curve based on the multi-scale fusion feature obtained by splicing the curvature feature, the morphological feature and the trend feature.

5. The finite element numerical simulation method for stress-strain curves according to claim 3, characterized in that, The fully connected layer includes: The first sub-layer is used to map the feature vector into 128-dimensional features and perform ReLU activation and Dropout regularization; The second sublayer is used to output the final strain of the compaction stage of the specimen by linear activation based on the output of the first sublayer.

6. The finite element numerical simulation method for stress-strain curves according to claim 2, characterized in that, The method further includes: Based on the evaluation index of the end strain of the compaction stage, the initial learning rate of the compaction stage identification model is determined to be 0.001, the batch size is 64, and training is stopped when the loss on the validation set no longer decreases after 15 consecutive training rounds. The evaluation metrics include mean absolute error, coefficient of determination, and relative error distribution; the learning rate is multiplied by 0.5 every 20 training epochs.

7. The finite element numerical simulation method for stress-strain curves according to claim 1, characterized in that, The step of performing uniaxial compression simulation on the preset initial three-dimensional model includes: fixing the lower end surface of the preset initial three-dimensional model, applying a velocity of 5 x 10 -8 meters per calculation step to the upper end surface of the preset initial three-dimensional model.

8. A finite element numerical simulation device for stress-strain curves, characterized in that, include: The curve acquisition module is used to obtain stress-strain curves based on uniaxial compression tests of specimens. The compaction stage determination module is used to determine the compaction stage of the sample based on the stress-strain curve. The model acquisition module is used to perform uniaxial compression simulation on the preset initial three-dimensional model, dynamically update the tangent modulus of the initial three-dimensional model in the compaction stage, and obtain a three-dimensional numerical model. The numerical simulation module is used to perform finite element numerical simulation of stress-strain curves based on the three-dimensional numerical model.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the finite element numerical simulation method for stress-strain curves as described in any one of claims 1 to 7.

10. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the finite element numerical simulation method for stress-strain curves as described in any one of claims 1 to 7.

Images