State variable guided support vector machine elastoplastic constitutive relation simulation method

Through the state variable-guided support vector machine method, the problem that non-time sequence ML model cannot represent the multi-valued hysteresis relationship and low computational efficiency is solved, and the complex multi-valued hysteresis relationship between force and deformation is realized, providing high efficiency The elastic-plastic constitutive relationship simulation method is suitable for building seismic prediction.

CN119047294APending Publication Date: 2024-11-29CHINA THREE GORGES UNIV
View PDF 0 Cites 2 Cited by

Patent Information

Application Number
CN202410960765.3
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-17
Publication Date
2024-11-29

AI Technical Summary

Technical Problem

The existing non-time sequence ML models cannot represent the multi-valued hysteresis relationship, and the traditional machine learning model has low computational efficiency and complex hyperparameter optimization, making it difficult to accurately predict the complex multi-valued hysteresis relationship between force and deformation.

Method used

Using the support vector machine (SVLS-SVM) method with state variable-guided guidance, the input variables are mapped to the state space by correcting the three-parameter hysteresis model, establishing a nonlinear single-valued function relationship between the state variable and the force or stress, simplifying hyperparameter optimization , improve the learning ability and computing efficiency of the model.

Benefits of technology

Accurate simulation of multi-value hysteresis relationships is realized, the complexity of hyperparameter optimization is reduced, the prediction accuracy and calculation efficiency of the model are improved, and it is suitable for hysteresis curve simulation of single-axis materials and column components, and has a wide range of Application prospects.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119047294A_ABST
    Figure CN119047294A_ABST
Patent Text Reader

Abstract

The invention discloses a state variable guided support vector machine elastoplastic constitutive relationship simulation method. The method comprises the following steps: 1, collecting elastoplastic constitutive model data sets of reinforced concrete columns of different failure types and different building materials; 2, dividing the hysteretic curve into a training set and a test set according to the time of the hysteretic curve; 3, mapping input variable displacement or strain to a state space by utilizing an operator for correcting the three-parameter hysteresis model to form a state variable capable of considering load path correlation; 4, based on an LS-SVM algorithm, establishing a nonlinear single-valued function relationship between the state variable and the force; 5, verifying the established state variable guided support vector machine elastoplastic constitutive relation simulation method, and comparing with the existing method; according to the method, a state function of a corrected three-parameter hysteresis model is utilized, a deformation input variable is mapped to a state space to form a state variable, and a single-valued function between force and the state variable can represent a complex multi-valued hysteresis relation between the force and deformation.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention relates to the technical field of structural earthquake resistance, and in particular to a state variable guided support vector machine elastoplastic constitutive relationship simulation method. Background Art

[0002] The elastic-plastic constitutive model of engineering materials or structural components is an indispensable part of the analysis of nonlinear seismic response of structures, and plays a key role in determining the accuracy and reliability of the analysis results. Since such models have complex nonlinear hysteresis characteristics with path correlation, it is a challenging problem to accurately construct the elastic-plastic constitutive model of materials or components. Hysteresis models based on smooth curves and multi-line types are often used to describe the elastic-plastic constitutive relations of materials or components. Since the elastic-plastic mechanical behaviors of different engineering materials and structural components are significantly different, it is impossible to use a unified hysteresis model to describe their elastic-plastic constitutive relations. For example, the smooth curve-based hysteresis model is often used to describe the elastic-plastic mechanical behavior of metal materials, while the multi-line hysteresis model is mostly used to describe the elastic-plastic mechanical behavior of reinforced concrete (RC) structures or components. Although the hysteresis curves of materials and components generated by the above hysteresis models can match the experimental data well, they still cannot fully reflect their complex experimental hysteresis behaviors.

[0003] At present, machine learning (ML) methods (such as recurrent neural networks, RNN) for time series data analysis use deformation (such as displacement or strain) as input and force (such as restoring force or stress) as output, which can accurately represent the elastic-plastic constitutive relationship of components or materials, and are used to replace the above-mentioned hysteresis models based on multi-line and smooth curves. The article "Deep learning predicts path-dependent plasticity" in "Proceedings of the National Academy of Sciences" 2019, Vol. 16, No. 52 uses the RNN model to replace the traditional RVE (Representative volume elements) technology to simulate the plastic behavior related to the load path, and discusses the combination of time domain and non-time domain features. "Theory-assisted small sample deep learning of elastic-plastic constitutive relations" in "Engineering Mechanics" 2023, Vol. 40, No. 9 proposed a small sample deep learning algorithm assisted by classical elastic-plastic mechanics theory, which can achieve good generalization performance under small data sample size. However, when building a model, the above method requires many hyperparameters to be adjusted (such as learning rate, number of iterations, number of hidden layers, number of neurons, etc.). Moreover, a non-convex optimization problem needs to be solved to obtain the model parameters, and the gradient descent algorithm is prone to fall into the local optimal solution during the solution process, and thus cannot guarantee the global optimal solution, making the parameter adjustment process time-consuming, and the adjusted optimal hyperparameters may only be applicable to the elastic-plastic constitutive model of a single component or material, and cannot be extended to other structural components or material types.

[0004] Elastic-plastic constitutive models are often used to describe complex nonlinear hysteresis characteristics with path correlation, while existing non-sequential ML models can simulate single-valued functions but cannot represent multi-valued hysteresis relationships. To address this problem, a sequential ML model can be used. Recurrent neural network (RNN) is a commonly used sequential ML model, which is widely used in sequential data analysis. Its underlying mathematical structure takes into account the sequential effects of data and can be used to simulate elastic-plastic constitutive relations. However, during the training process of the gradient descent optimization algorithm, as the sequence length increases, the gradient may disappear or explode when RNN backtracks in time to calculate the gradient, making the learning process very difficult or unstable. Due to the problem of vanishing gradients, RNN is difficult to capture long-term dependencies in practical applications. It is easy to fall into a local optimal solution, and thus cannot guarantee a global optimal solution. Moreover, when constructing an RNN model, there are many hyperparameters that need to be adjusted and the parameter adjustment process is time-consuming. These hyperparameters include the size of the hidden layer, the learning rate, and the number of iterations, which have a crucial impact on the performance of the RNN model. Therefore, it is urgent to provide a complex multi-valued hysteresis relationship that can accurately predict the relationship between force and deformation. Summary of the invention

[0005] The technical problem to be solved by the present invention is to provide a state variable guided support vector machine elastic-plastic constitutive relationship simulation method to solve the two key problems that the existing non-time series ML models can simulate single-valued functions but cannot represent multi-valued hysteresis relationships, and the traditional machine learning models have low computational efficiency and complex hyperparameter optimization.

[0006] In order to solve the above technical problems, the technical solution adopted by the present invention is a state variable guided support vector machine elastic-plastic constitutive relationship simulation method, comprising the following steps: Step 1: Collect test data of reinforced concrete columns with different failure types and elastic-plastic constitutive models of different building materials, including symmetric and asymmetric hysteresis curves of hysteresis loops with different geometric shapes; Step 2: Divide all data into training set and test set according to the time characteristics of hysteresis loops of different geometric shapes; Step 3: Use the operator of the modified three-parameter hysteresis model, i.e., the state function, to map the input variable displacement or strain to the state space to form a state variable that can take into account the correlation of the load path; Step 4: Based on the LS-SVM algorithm, a nonlinear single-valued function relationship between the state variable and the force or stress is established to establish the SVLS-SVM model; Step 5: Based on the test set data, the prediction performance of the state variable guided support vector machine elastoplastic constitutive relationship simulation method is verified and compared with the existing methods.

[0007] In a preferred solution, the test data of reinforced concrete columns of different failure types in step 1 include bending failure, shear failure and bending-shear failure, and the elastic-plastic constitutive models of different building materials include steel bars and concrete materials on Opensees (Open Systemfor Earthquake Engineering Simulation).

[0008] In a preferred solution, the training set and the test set established in step 2 are divided in a ratio of 1:1 according to the timeliness of the hysteresis curve.

[0009] In a preferred embodiment, the input state variables in step 3 specifically include: displacement / strain, loading-unloading indicator, accumulated energy, positive displacement / strain at zero load in a single loading-unloading closed-loop process, negative displacement / strain at zero load in a single loading-unloading closed-loop process, maximum positive displacement / strain in a single loading-unloading closed-loop process, and maximum negative displacement / strain in a single loading-unloading closed-loop process.

[0010] In the preferred solution, the theoretical basis for calculating the state variables in Step 3 is to use the operator of the modified three-parameter hysteresis model, i.e., the state function, to map the input variable displacement or strain to the state space to form a state variable that can consider the correlation of the load path. The mathematical principle is expressed as follows: By introducing a point in the positive direction on the extension line of the initial stiffness and a point in the negative direction , which effectively represents the degradation of unloading stiffness, where is the unloading stiffness degradation parameter, from which the unloading stiffness in the positive and negative directions of loading and unloading can be obtained and The calculation formula is as follows: (1) (2) In the formula, Specify the degree of stiffness degradation (this value is the hysteresis parameter), The smaller the value, the more serious the stiffness degradation. represents the degenerate displacement in the negative direction, represents the degenerate force in the negative direction, represents the degenerate displacement in the positive direction, represents the degenerate force in the positive direction, Indicates the current load step i The target maximum point in the positive direction, express The displacement of the point in the positive direction, express The force in the positive direction of the point; When loading and unloading in positive and negative directions, the state variable accumulates energy The calculation formula is as follows: (3) In the formula, is the restoring force of the previous load step, is the displacement increment, is the accumulated energy of the current load step, is the accumulated energy of the previous load step, The restoring forces or stresses from the previous load step, is the restoring force or stress of the current load step; Combined with formula (1), the negative hysteresis displacement is obtained The calculation formula is as follows: (4) In the formula, is the displacement of the previous loading step; When the maximum or minimum deformation exceeds the yield deformation in the positive or negative direction of the current load cycle, cyclic strength degradation occurs. The deformation exceeds the yield deformation in the positive direction during the cycle, so the reloading path in the positive direction of the next load cycle is guided toward the new maximum point due to strength degradation. Combined with formula (4), the cumulative hysteresis energy that controls the cyclic strength degradation can be obtained: The calculation formula is as follows: (5) In the formula, is the hysteresis parameter, which controls the degree of cycle strength degradation related to the cumulative energy dissipation, is the total energy; By combining equation (4) and equation (5), the maximum positive displacement of the state variable in a single loading and unloading closed-loop process can be obtained: The calculation formula is as follows: (6).

[0011] In the preferred solution, the establishment of the SVLS-SVM model in Step 4 is to establish a nonlinear single-valued function relationship between the state variable and the force or stress, which is used to represent the complex multi-valued hysteresis relationship between deformation and load: (7) In the formula, , , is the kernel matrix composed of kernel functions, b is the model parameter in the dual space, here , which fully reflects the test data The nonlinear mapping relationship between them realizes that the single-valued function between force and state variables can express the complex multi-valued hysteresis relationship between force and deformation.

[0012] In the preferred solution, the prediction performance evaluation in Step 5 is to compare the SVLS-SVM model with common machine learning methods, which are mainly recurrent neural networks (RNNs), by analyzing the determination coefficients of each model on the training set and the test set. , the mean absolute error (MAE) of the test set, and the model training time are used to comprehensively evaluate the training efficiency and prediction accuracy of these models. Through comparative analysis with the prior art, the application value of the present invention in the field of hysteresis curve prediction is confirmed.

[0013] The state variable guided support vector machine elastic-plastic constitutive relationship simulation method provided by the present invention has the following beneficial effects: 1. The SVLS-SVM model proposed in the present invention only needs to optimize two hyperparameters, and the optimized hyperparameters are applicable to the hysteresis curve simulation of all uniaxial materials and column components verified in this paper, without the need to repeatedly optimize the hyperparameters for each uniaxial material or column component; 2. The SVLS-SVM model has strong learning ability and can accurately learn symmetric and asymmetric hysteresis curves with different geometric shapes, and can ensure the global optimal solution. Although the RNN model can accurately characterize these hysteresis curves, it is easy to fall into the local optimal solution, resulting in low simulation accuracy; 3. The present invention compares the SVLS-SVM model with common machine learning methods and verifies that the training process of the SVLS-SVM model is equivalent to solving a set of linear equations, which can directly obtain the parameters of the prediction model. The training process of the RNN model needs to be achieved by optimizing the loss function and iterating the model parameters multiple times; 4. The SVLS-SVM model proposed in the present invention does not rely on large-scale data sets during training, and only requires a small number of samples to achieve high simulation accuracy, which makes it more flexible and efficient in practical applications; in addition, the SVLS-SVM model also has good robustness and can cope with challenges such as data noise and outliers; 5. The SVLS-SVM model proposed in the present invention is not only applicable to the hysteresis curve simulation of uniaxial materials and column components, but can also be extended to the constitutive relationship simulation of multiaxial materials and complex structural systems, and has a wide range of application prospects; 6. The SVLS-SVM model proposed in the present invention has a clear mathematical explanation and theoretical basis in the simulation process, making its simulation results easier to understand and analyze; at the same time, the model also has good interpretability and generalizability, providing a powerful tool support for research in the fields of materials science and civil engineering. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] The present invention will be further described below in conjunction with the accompanying drawings and implementation examples: Figure 1 It is the technical roadmap of the present invention; Figure 2 This is the main working principle of the SVLS-SVM model of the present invention. DETAILED DESCRIPTION

[0015] The technical solution of the present invention is further described below in conjunction with the accompanying drawings and embodiments: Example 1 like Figure 1~2 As shown in FIG. 1 , the state variable guided support vector machine elastic-plastic constitutive relationship simulation method includes the following steps: Step 1: Collect test data of reinforced concrete columns with different failure types and elastic-plastic constitutive models of different building materials, including symmetric and asymmetric hysteresis curves of hysteresis loops with different geometric shapes; Step 2: Divide all data into training set and test set according to the time characteristics of hysteresis loops of different geometric shapes; Step 3: Use the operator of the modified three-parameter hysteresis model, i.e., the state function, to map the input variable displacement or strain to the state space to form a state variable that can take into account the correlation of the load path; Step 4: Based on the LS-SVM algorithm, a nonlinear single-valued function relationship between the state variable and the force or stress is established to establish the SVLS-SVM model; Step 5: Based on the test set data, the prediction performance of the state variable guided support vector machine elastoplastic constitutive relationship simulation method is verified and compared with the existing methods.

[0016] In a preferred solution, the test data of reinforced concrete columns of different failure types in step 1 include bending failure, shear failure and bending-shear failure, and the elastic-plastic constitutive models of different building materials include steel bars and concrete materials on Opensees.

[0017] In a preferred solution, the training set and the test set established in step 2 are divided in a ratio of 1:1 according to the timeliness of the hysteresis curve.

[0018] In a preferred embodiment, the input state variables in step 3 specifically include: displacement / strain, loading-unloading indicator, accumulated energy, positive displacement / strain at zero load in a single loading-unloading closed-loop process, negative displacement / strain at zero load in a single loading-unloading closed-loop process, maximum positive displacement / strain in a single loading-unloading closed-loop process, and maximum negative displacement / strain in a single loading-unloading closed-loop process.

[0019] In the preferred solution, the theoretical basis for calculating the state variables in Step 3 is to use the operator of the modified three-parameter hysteresis model, i.e., the state function, to map the input variable displacement or strain to the state space to form a state variable that can consider the correlation of the load path. The mathematical principle is expressed as follows: By introducing a point in the positive direction on the extension line of the initial stiffness and a point in the negative direction , which effectively represents the degradation of unloading stiffness, where is the unloading stiffness degradation parameter, from which the unloading stiffness in the positive and negative directions of loading and unloading can be obtained and The calculation formula is as follows: (1) (2) In the formula, Specify the degree of stiffness degradation (this value is the hysteresis parameter), The smaller the value, the more serious the stiffness degradation. represents the degenerate displacement in the negative direction, represents the degenerate force in the negative direction, represents the degenerate displacement in the positive direction, represents the degenerate force in the positive direction, Indicates the current load step i The target maximum point in the positive direction, express The displacement of the point in the positive direction, express The force in the positive direction of the point; When loading and unloading in positive and negative directions, the state variable accumulates energy The calculation formula is as follows: (3) In the formula, is the restoring force of the previous load step, is the displacement increment, is the accumulated energy of the current load step, is the accumulated energy of the previous load step, The restoring forces or stresses from the previous load step, is the restoring force or stress of the current load step; Combined with formula (1), the negative hysteresis displacement is obtained The calculation formula is as follows: (4) In the formula, is the displacement of the previous loading step; When the maximum or minimum deformation exceeds the yield deformation in the positive or negative direction of the current load cycle, cyclic strength degradation occurs. The deformation exceeds the yield deformation in the positive direction during the cycle, so the reloading path in the positive direction of the next load cycle is guided toward the new maximum point due to strength degradation. Combined with formula (4), the cumulative hysteresis energy that controls the cyclic strength degradation can be obtained: The calculation formula is as follows: (5) In the formula, is the hysteresis parameter, which controls the degree of cycle strength degradation related to the cumulative energy dissipation, is the total energy; By combining equation (4) and equation (5), the maximum positive displacement of the state variable in a single loading and unloading closed-loop process can be obtained: The calculation formula is as follows: (6).

[0020] In the preferred solution, the establishment of the SVLS-SVM model in Step 4 is to establish a nonlinear single-valued function relationship between the state variable and the force or stress, which is used to represent the complex multi-valued hysteresis relationship between deformation and load: (7) In the formula, , , is the kernel matrix composed of kernel functions, b is the model parameter in the dual space, here , which fully reflects the test data The nonlinear mapping relationship between them realizes that the single-valued function between force and state variables can express the complex multi-valued hysteresis relationship between force and deformation.

[0021] In the preferred solution, the prediction performance evaluation in Step 5 is to compare the SVLS-SVM model with common machine learning methods, which are mainly recurrent neural networks (RNNs), by analyzing the determination coefficients of each model on the training set and the test set. , the mean absolute error (MAE) of the test set, and the model training time are used to comprehensively evaluate the training efficiency and prediction accuracy of these models. Through comparative analysis with the prior art, the application value of the present invention in the field of hysteresis curve prediction is confirmed.

[0022] Example 2 In another preferred embodiment, based on the above embodiment 1, the verification and performance evaluation in Step 5 is to compare the SVLS-SVM model with common machine learning methods, which are recurrent neural networks (RNNs). The training efficiency and prediction accuracy of these models are comprehensively evaluated by comparing the R squared, the MAE (Mean Absolute Error) of the test set, and the model training time. The application value of the present invention in the field of hysteresis curve prediction is confirmed by comparative analysis with the prior art. The specific results are shown in Table 1, which is a performance comparison of SVLS-SVM and RNN models based on different RC (Reinforced Concrete) column and engineering material data: Table 1 Performance comparison of SVLS-SVM and RNN models based on different RC column and engineering material data

[0023] Table 1 shows the performance indicators of the SVLS-SVM model and the RNN model in more detail. In Table 1, the SVLS-SVM model shows excellent performance on both the training set and the test set. The values ​​of both reached 0.99, and the training time of the model was less than 1 second, which shows its efficient training ability. The value is slightly lower than that of SVLS-SVM, but its training time is significantly longer, up to 878.115 seconds. This shows that there is a clear gap in prediction accuracy, and the overall performance of RNN is far inferior to the prediction performance of the SVLS-SVM model.

[0024] In summary, the state variable guided support vector machine elastic-plastic constitutive relationship simulation method provided by the present invention is used to solve the two key problems that the existing non-time series ML models can simulate single-valued functions but cannot represent multi-valued hysteresis relationships, and the traditional machine learning models have low computational efficiency and complex hyperparameter optimization. The computational efficiency and prediction performance of the elastic-plastic constitutive model of different types of reinforced concrete column test data and different building materials established by the comparative analysis with the recurrent neural network (RNN) method are verified, which has important reference significance for promoting the development of building seismic prediction methods; at the same time, the present invention also proposes a new elastic-plastic constitutive relationship simulation method based on SVLS-SVM, which can effectively simulate multi-valued hysteresis relationships, improve the prediction accuracy and computational efficiency of the model, and provide new ideas and methods for building seismic prediction; in practical applications, the method can be widely used in the evaluation and prediction of seismic performance of building structures, and provide a scientific reference basis for engineering design and construction; in addition, the proposal of this method also opens up a new direction for the application of machine learning in the field of civil engineering, and has broad application prospects and important academic value.

Claims

1. The state variable guided support vector machine elastic-plastic constitutive relationship simulation method is characterized by: The method comprises the following steps: Step 1: Collect test data of reinforced concrete columns with different failure types and elastic-plastic constitutive models of different building materials, including symmetric and asymmetric hysteresis curves of hysteresis loops with different geometric shapes; Step 2: Divide all data into training set and test set according to the time characteristics of hysteresis loops of different geometric shapes; Step 3: Use the operator of the modified three-parameter hysteresis model, i.e., the state function, to map the input variable displacement or strain to the state space to form a state variable that can take into account the correlation of the load path; Step 4: Based on the LS-SVM algorithm, a nonlinear single-valued function relationship between the state variable and the force or stress is established to establish the SVLS-SVM model; Step 5: Based on the test set data, the prediction performance of the state variable guided support vector machine elastoplastic constitutive relationship simulation method is verified and compared with the existing methods.

2. The state variable guided support vector machine elastic-plastic constitutive relationship simulation method according to claim 1, characterized in that: The test data of reinforced concrete columns with different types of failure in step 1 include bending failure, shear failure and bending-shear failure, and the elastic-plastic constitutive models of different building materials include steel bars and concrete materials on Opensees.

3. The state variable guided support vector machine elastic-plastic constitutive relationship simulation method according to claim 1, characterized in that: The training set and test set established in step 2 are divided in a 1:1 ratio according to the timeliness of the hysteresis curve.

4. The state variable guided support vector machine elastic-plastic constitutive relationship simulation method according to claim 1, characterized in that: The input state variables in step 3 specifically include: displacement / strain, loading-unloading indicator, accumulated energy, positive displacement / strain at zero load in a single loading-unloading closed-loop process, negative displacement / strain at zero load in a single loading-unloading closed-loop process, maximum positive displacement / strain in a single loading-unloading closed-loop process, and maximum negative displacement / strain in a single loading-unloading closed-loop process.

5. The state variable guided support vector machine elastic-plastic constitutive relationship simulation method according to claim 1, characterized in that: The theoretical basis for calculating the state variables in Step 3 is to use the operator of the modified three-parameter hysteresis model, i.e., the state function, to map the input variable displacement or strain to the state space to form a state variable that can consider the correlation of the load path. The mathematical principle is expressed as follows: By introducing a point in the positive direction on the extension line of the initial stiffness and a point in the negative direction , which effectively represents the degradation of unloading stiffness, where is the unloading stiffness degradation parameter, from which the unloading stiffness in the positive and negative directions of loading and unloading can be obtained and The calculation formula is as follows: (1) (2) In the formula, Specify the degree of stiffness degradation (this value is the hysteresis parameter), The smaller the value, the more serious the stiffness degradation. represents the degenerate displacement in the negative direction, represents the degenerate force in the negative direction, represents the degenerate displacement in the positive direction, represents the degenerate force in the positive direction, Indicates the current load step i The target maximum point in the positive direction, express The displacement of the point in the positive direction, express The force in the positive direction of the point; When loading and unloading in positive and negative directions, the state variable accumulates energy The calculation formula is as follows: (3) In the formula, is the restoring force of the previous load step, is the displacement increment, is the accumulated energy of the current load step, is the accumulated energy of the previous load step, The restoring forces or stresses from the previous load step, is the restoring force or stress of the current load step; Combined with formula (1), the negative hysteresis displacement is obtained The calculation formula is as follows: (4) In the formula, is the displacement of the previous loading step; When the maximum or minimum deformation exceeds the yield deformation in the positive or negative direction of the current load cycle, cyclic strength degradation occurs. The deformation exceeds the yield deformation in the positive direction during the cycle, so the reloading path in the positive direction of the next load cycle is guided toward the new maximum point due to strength degradation. Combined with formula (4), the cumulative hysteresis energy that controls the cyclic strength degradation can be obtained: The calculation formula is as follows: (5) In the formula, is the hysteresis parameter, which controls the degree of cycle strength degradation related to the cumulative energy dissipation, is the total energy; By combining equation (4) and equation (5), the maximum positive displacement of the state variable in a single loading and unloading closed-loop process can be obtained: The calculation formula is as follows: (6)。 6. The state variable guided support vector machine elastic-plastic constitutive relationship simulation method according to claim 1, characterized in that: The establishment of the SVLS-SVM model in Step 4 is to establish a nonlinear single-valued function relationship between the state variable and the force or stress, which is used to express the complex multi-valued hysteresis relationship between deformation and load: (7) In the formula, , , is the kernel matrix composed of kernel functions, and b is the model parameter in the dual space.

7. The state variable guided support vector machine elastic-plastic constitutive relationship simulation method according to claim 1, characterized in that: The prediction performance evaluation in Step 5 is to compare the SVLS-SVM model with common machine learning methods by analyzing the determination coefficient of each model on the training set and test set. , the mean absolute error (MAE) of the test set, and the model training time are used to comprehensively evaluate the training efficiency and prediction accuracy of these models.

Citation Information

Cited By

  • Component hysteretic curve prediction method based on multi-dimensional means, terminal and medium

    CN122197662A

  • A multi-dimensional method for predicting component hysteresis curves, terminals, and media.

    CN122197662B