A method and system for predicting fatigue life using a custom rubber constitutive model

By developing a custom constitutive model for fatigue life prediction, this method addresses the issues of limited model selection and insufficient accuracy in fatigue life prediction of rubber materials. It achieves high-precision fatigue life prediction and improves the efficiency of batch analysis, while providing various accuracy evaluation schemes for constitutive models.

CN123120155APending Publication Date: 2026-11-17BEIJING UNIV OF CHEM TECH +1
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611565901.4
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-10-10
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

In existing technologies, fatigue life prediction for rubber materials suffers from problems such as a limited range of hyperelastic constitutive models, insufficient correlation between model fitting accuracy and fatigue life prediction results, and a lack of standardized model selection procedures, making it difficult to achieve high-precision fatigue life prediction.

Method used

A method for predicting fatigue life using a custom rubber constitutive model is provided. By acquiring mechanical experimental data and fatigue crack propagation rate of rubber materials, parameter fitting is performed based on multiple hyperelastic constitutive models, a UHYPER subroutine is written, an Endurica solver configuration file is generated, fatigue simulation batch processing at multiple load levels is performed, and the normalized area method is used to evaluate the model accuracy.

Benefits of technology

It expands the range of constitutive models to be selected, improves the accuracy of fatigue life prediction and the efficiency of batch analysis, and provides a basis for quantitatively evaluating the fatigue life range prediction accuracy of various hyperelastic constitutive models, thus providing a basis for subsequent complex multiaxial fatigue analysis.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123120155A_ABST
    Figure CN123120155A_ABST
Patent Text Reader

Abstract

This invention provides a method and system for predicting the fatigue life of a custom rubber constitutive model, belonging to the field of fatigue life prediction for rubber materials. The method includes: acquiring mechanical experimental data, fatigue crack propagation rate, and tensile fatigue life data of rubber materials under at least one deformation mode; fitting hyperelastic constitutive model parameters and writing a UHYPER subroutine; fitting a fatigue crack propagation model based on the fatigue crack propagation rate and incorporating the material model data into the solver's required configuration file; performing batch processing of fatigue simulations with multiple constitutive models and multiple load levels, associating the constitutive model source file with the load configuration file, calculating the fatigue life under each load sequentially, and archiving the results; and using the normalized area method to quantitatively evaluate the prediction accuracy of the fatigue life range of different hyperelastic constitutive models. This invention overcomes the limitations of traditional finite element software, achieving efficient batch processing and quantitative accuracy evaluation of custom constitutive models.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention belongs to the field of fatigue life prediction of rubber materials, and specifically relates to a method and system for predicting fatigue life using a custom rubber constitutive model. Background Technology

[0002] Rubber materials possess excellent physicochemical properties, including high elasticity, viscoelastic damping, sealing performance, and electrical insulation, making them indispensable in various modern scientific and engineering fields. These materials are widely used in critical components such as tires, air springs, conveyor belts, seals, and vibration damping supports. However, under long-term cyclic loading, rubber materials are prone to fatigue failure. During service, microcracks within the material gradually propagate, eventually causing the component to lose its macroscopic load-bearing capacity. Fatigue failure of rubber materials can cause significant economic losses and serious safety hazards. Therefore, achieving high-precision fatigue life prediction for rubber materials has become a core issue that urgently needs to be addressed by both academia and industry.

[0003] Crack propagation methods based on fracture mechanics have been widely used in fatigue life analysis of various rubber products. The core of this method lies in the accurate quantification of crack driving force. Griffith proposed the concept of energy release rate (also known as tear energy), defined as T = -d(UV) / dA, where U is the potential energy of the crack configuration, V is the work done by the external force during crack propagation, and A is the crack surface area. Since tear energy is directly related to mechanical response, and mechanical response is usually characterized by hyperelastic constitutive models, the accuracy of hyperelastic constitutive models is an important prerequisite for accurately predicting stress, strain, tear energy, and fatigue life. However, there are currently nearly a hundred hyperelastic constitutive models, and how to select a suitable model and complete numerical calculations remains an urgent engineering problem to be solved.

[0004] Although hyperelastic constitutive models play a crucial role in fatigue life prediction, systematic comparative studies on the performance of various models are still limited. Existing literature mostly focuses on a few classic models. For example, Mujdeci et al. evaluated the applicability of Yeoh, Mooney-Rivlin, Ogden, and Neo-Hookean models in tire finite element analysis, pointing out that model applicability is related to strain level. Zhang et al. studied flexible joint elastomers, concluding that the Ogden model outperforms the Mooney-Rivlin model under complex loading conditions. Peng et al. derived the integral formula for the energy density of rubber cracks under finite deformation based on the Ogden model, finding that the numerical differences between higher-order Ogden models (order N=3-6) are small. Kim et al. further verified the advantage of the third-order Ogden model in the large deformation problem of neoprene rubber, while pointing out that the applicable strain range of the Neo-Hookean and Mooney-Rivlin models is limited. Although the above studies have demonstrated the importance of model selection, they generally have three core shortcomings: (1) Limited research scope: the candidate model library has a narrow coverage and cannot fully cover the current rich variety of hyperelastic constitutive theory systems; (2) Insufficient correlation: the quantitative mapping relationship between model fitting accuracy and fatigue life prediction results is still unclear; (3) Weak guidance: there is a lack of standardized model selection process for engineering applications.

[0005] Furthermore, nearly a hundred hyperelastic constitutive models have been developed. However, due to the difficulty of numerical calculation and the lack of model selection criteria, many models face challenges in practical engineering applications. Researchers and engineers mostly still choose classic models such as Ogden, Yeoh, and Mooney-Rivlin. This situation makes it difficult to select the most suitable hyperelastic constitutive model based on material properties, thus limiting the high-precision prediction of rubber fatigue life. Therefore, there is an urgent need in this field to construct a fatigue life analysis method compatible with multiple hyperelastic constitutive models, laying a solid theoretical and technical foundation for high-fidelity fatigue life assessment. Summary of the Invention

[0006] The present invention aims to solve the problems of the prior art and provide a method for predicting the fatigue life of rubber materials based on a custom hyperelastic constitutive model.

[0007] This invention is achieved through the following technical solution: A first aspect of the present invention provides a method for predicting the fatigue life of a custom rubber constitutive model, characterized in that the method comprises: Step 1: Obtain mechanical experimental data, fatigue crack propagation rate and tensile fatigue life data of rubber materials under at least one deformation mode; Step 2: Based on the mechanical experimental data, perform parameter fitting on the selected hyperelastic constitutive models to obtain the hyperelastic constitutive model parameters corresponding to the rubber material, and write UHYPER subroutine code according to the strain energy density function of the hyperelastic constitutive model and its first derivative, second derivative and mixed partial derivative with respect to strain invariants. Step 3: Fit a fatigue crack propagation model based on the fatigue crack propagation rate, and compile the material model data into the configuration file required by the Endurica solver; Step 4: Perform batch processing of fatigue simulations for multiple hyperelastic constitutive models and multiple load levels, including: obtaining multiple constitutive model source files and their corresponding fatigue analysis configuration files for multiple load levels; establishing the association between the source files and configuration files and performing integrity verification; compiling the constitutive model source files that pass the verification; calculating the fatigue life under each load and archiving the results.

[0008] Step 5: Based on the tensile fatigue test data of rubber materials and the fatigue life prediction range of each hyperelastic constitutive model at various strain levels, the normalized area method is used to quantitatively evaluate the fatigue life range prediction accuracy of different hyperelastic constitutive models.

[0009] Furthermore, in step 2, the parameter fitting employs the least squares method, and the fitting accuracy is determined by the coefficient of determination. The formula is as follows:

[0010] in, For the i-th experimental stress value, For the i-th predicted nominal stress value, denoted as the average experimental stress, and n as the total number of data points.

[0011] Furthermore, the configuration file in step 3 includes: Output control information, used to specify the output items required for fatigue life calculation; Material constitutive information, used to specify the hyperelastic constitutive behavior of the material as UHYPER; Crack precursor size information, including initial crack size and terminal crack size; Fatigue crack propagation model information, used to specify the type and parameters of the fatigue crack propagation model; The loading history data file is used to provide strain history loading data.

[0012] Furthermore, the operation of compiling the source files of the constitutive models that have passed the verification in step 4 includes: determining whether the number of fatigue analysis configuration files corresponding to each constitutive model has reached a preset number; performing subsequent processing only on constitutive models that have reached the preset number, and skipping constitutive models that have not reached the preset number and recording abnormal information. The fatigue analysis configuration file adopts the naming method of constitutive model identifier + load level identifier; the constitutive model source file corresponding to the same constitutive model and multiple fatigue analysis configuration files have the same constitutive model identifier, and different fatigue analysis configuration files are distinguished by different load level identifiers.

[0013] Furthermore, the preset quantity is 3, which corresponds to the fatigue life calculation under three preset strain levels.

[0014] Furthermore, the operation of compiling the source files of the constitutive model that have passed the verification in step 4 includes: calling the compiler to compile the source files of the constitutive model into a dynamic link library that can be called by the fatigue analysis program; and, each source file that has passed the verification is compiled only once, so that it can be reused for multiple load level calculations corresponding to the same constitutive model. The operation of sequentially calculating fatigue life under each load includes: detecting the calculation execution status of each fatigue analysis configuration file; continuing to execute the calculation of the next load level when the current load level calculation is successful; stopping the subsequent calculation of the current constitutive model file group and recording the failure information when any load level calculation fails. The archiving calculation operation includes: creating a result directory according to the constitutive model identifier, archiving the constitutive model source file, fatigue analysis configuration file, compiled generated file and fatigue life calculation result file to the result directory, and recording abnormal information, which includes at least one of compilation failure information, calculation failure information and configuration file missing information.

[0015] Furthermore, step 5, which uses the normalized area method to quantitatively evaluate the fatigue life range prediction accuracy of different hyperelastic constitutive models, includes: For each strain level, the predicted fatigue life range and the experimental fatigue life range at that strain level are determined respectively. Using experimental fatigue life as the abscissa and predicted fatigue life as the ordinate, the area enclosed between the line segment representing the range of predicted fatigue life and the range of experimental fatigue life and the ideal line y=x is taken as the original area under this strain level. The original area is normalized to obtain the normalized area at this strain level; Evaluation indices are calculated based on normalized areas at multiple strain levels. These indices are used to characterize the accuracy of fatigue life range prediction for the hyperelastic constitutive model.

[0016] Furthermore, the original area is calculated in the following way: Define the signed longitudinal deviation dA of the two endpoints of a line segment relative to the ideal line y=x. - dB= - When dA•dB≥0, the original area is calculated as a trapezoid; when dA•dB<0, the original area is calculated as two triangles. The normalized area is calculated using the following formula:

[0017] in, Normalized area under specific strain and These represent the maximum and minimum values ​​of the experimental fatigue life, respectively. The original area is denoted as .

[0018] A second aspect of the present invention provides a custom hyperelastic constitutive rubber fatigue life prediction system, the system comprising: The data acquisition module is used to acquire mechanical experimental data, fatigue crack propagation rate data, and tensile fatigue life data of rubber materials under at least one deformation mode and at least three strain levels. The constitutive subroutine module is used to perform parameter fitting on various hyperelastic constitutive models based on strain invariants based on the mechanical experimental data, and to write UHYPER subroutines for each constitutive model according to its strain energy density function and the first, second, and mixed partial derivatives with respect to the first strain invariant, the second strain invariant, and the volume ratio, and to establish the mapping between theoretical variables and UHYPER interface variables. The configuration file generation module is used to fit the Lake-Lindley model and the Thomas model based on the fatigue crack propagation rate data and generate the configuration file used by the Endurica solver. The batch processing scheduling module is used to associate each constitutive model source file with fatigue analysis configuration files for multiple load levels according to the file identifier, and to perform integrity verification on the associated file group; it compiles the source files with a preset number of load configuration files into a reusable dynamic link library for the fatigue solver, and sequentially calls the fatigue solver to calculate fatigue life according to the load level. The results archiving module is used to create a results directory according to the constitutive model identifier, archive source files, configuration files, compiled files and calculation results, and record compilation status, calculation status and exception information; The accuracy assessment module is used to obtain the predicted lifetime range of each constitutive model at various strain levels based on the maximum and minimum values ​​of the crack precursor size. Combined with the experimental lifetime range, it calculates the original area enclosed by the experimental-prediction line segment and the ideal line y=x. The normalized area of ​​each strain level and the average normalized area and standard deviation of multiple strain levels are calculated to evaluate the accuracy of fatigue life range prediction for different constitutive models.

[0019] Furthermore, in the batch processing scheduling module, the integrity verification operation includes: determining whether the number of fatigue analysis configuration files corresponding to each constitutive model has reached a preset number; only performing compilation and calculation on constitutive models that have reached the preset number, and skipping constitutive models that have not reached the preset number and recording abnormal information.

[0020] The beneficial effects of this invention are: Expanding the range of constitutive models: Breaking through the limitation of traditional finite element software that only has a few classic models built-in, a custom hyperelastic constitutive model is introduced into rubber fatigue life analysis, providing technical support for improving prediction accuracy.

[0021] Improve batch analysis efficiency: Through the automatic association between constitutive model source files and multi-load configuration files and the compilation result reuse mechanism, each constitutive model only needs to be compiled once to complete the continuous calculation of multiple load levels, which greatly reduces repeated compilation, manual configuration and result sorting operations, and reduces the risk of file mismatch.

[0022] Quantitative evaluation of model accuracy: The normalized area method is used to quantitatively evaluate the life range prediction accuracy of various hyperelastic constitutive models under tensile fatigue, providing a basis for model selection in subsequent complex multiaxial fatigue analysis. Attached Figure Description

[0023] Figure 1 This is a flowchart of a method according to an embodiment of the present invention; Figure 2 The tensile fatigue data are from Embodiment 6 of the present invention; Figure 3 The stress-strain data output by the UHYPER subroutine in the Endurica program in Embodiment 6 of the present invention are the fitting data of the hyperelastic constitutive model. Figure 4 The R² ranking of the 36 models fitted to the stress-strain curves of natural rubber / butadiene rubber composite materials in Example 6 of this invention; Figure 5 This is a comparison of the fatigue life prediction range of the Gent model calibrated based on the test data of natural rubber / butadiene rubber composite material in Example 6 of the present invention with the test data under different strain levels of tensile fatigue. Figure 6The average normalized area and standard deviation of 36 hyperelastic constitutive models obtained by fitting experimental data of natural rubber / butadiene rubber composite materials in Example 6 of this invention; Figure 7 This represents the average normalized area of ​​the 36 hyperelastic constitutive models in Embodiment 6 of the present invention. Detailed Implementation

[0024] The present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Addressing the challenge of applying and rationally selecting various hyperelastic constitutive models in fatigue life prediction in the prior art, the present invention provides a method for predicting the fatigue life of rubber materials based on a custom hyperelastic constitutive model, such as... Figure 1 As shown, it includes the following steps: Step 1: Obtain mechanical experimental data, fatigue crack propagation rate and tensile fatigue life data of rubber materials under at least one deformation mode; Step 2: Based on the mechanical experimental data, perform parameter fitting on the selected hyperelastic constitutive model to obtain the hyperelastic constitutive model parameters corresponding to the rubber material, and write user-defined UHYPER subroutine code according to the strain energy density function of the hyperelastic constitutive model and its first derivative, second derivative and mixed partial derivative with respect to strain invariants. Step 3: Fit the corresponding fatigue crack propagation model based on the fatigue crack propagation rate, and compile the fatigue crack propagation model parameters into the key files required by the Endurica solver: ki.json and .db files; Step 4: Perform batch processing of fatigue simulations for multiple hyperelastic constitutive models and multiple load levels. Obtain multiple Fortran source files containing material constitutive relations and corresponding model parameters, and establish associations between each constitutive model source file and fatigue analysis configuration files corresponding to multiple load levels based on file identifiers. Verify the integrity of the associated files, and only perform subsequent processing on constitutive models that simultaneously have a preset number of load configuration files. Call the compiler to compile the constitutive model source files that have passed the integrity verification into dynamic link libraries that can be called by the fatigue analysis program; after successful compilation, input the configuration files for different load levels into the fatigue analysis program in sequence to calculate the fatigue life under the corresponding load conditions. After all load conditions are calculated, create a results directory according to the constitutive model identifier, archive the constitutive model files, load configuration files, and fatigue calculation results to the corresponding results directory, and record the compilation status, calculation status, and error information of each constitutive model.

[0025] Step 5: Due to the discreteness of the size of the crack precursor, the fatigue life of rubber materials fluctuates greatly. Based on the tensile fatigue test data of rubber materials and the fatigue life prediction range of each hyperelastic constitutive model under various strains, the normalized area method is used to quantitatively evaluate the fatigue life range prediction accuracy of different hyperelastic constitutive models.

[0026] Example 1 This embodiment provides a detailed description of step 1. Step 1 involves obtaining mechanical experimental data of the rubber material under uniaxial tensile loading, fatigue crack propagation rates at different tear energies, and tensile fatigue life data at three strain levels. These three strain levels correspond to 95%, 131%, and 152%, respectively.

[0027] The acquisition of stress-strain experimental data for rubber materials under various deformation modes includes: uniaxial tension (UT), uniaxial compression (UC), planar tension (PT), and iso-biaxial tension (ET), i.e., at least one deformation mode is selected from one or more of uniaxial tension (UT), uniaxial compression (UC), planar tension (PT), and iso-biaxial tension (ET); when the mechanical experimental data only includes uniaxial tension data, in step 2, when fitting the hyperelastic constitutive model parameters, at least one material parameter is fixed or a literature value is used, or other deformation mode data are combined to improve the uniqueness of parameter fitting; Preferably, stress-strain experimental data are obtained under at least two deformation modes to more accurately determine the parameters of the hyperelastic constitutive model; The fatigue crack propagation data includes the fatigue crack propagation rate at a wide tear energy, the critical tear energy, and / or the fatigue threshold; the tensile fatigue life data requires at least three strains and at least seven parallel specimens at each strain.

[0028] Mechanical experimental data include at least one of uniaxial tensile test data, planar tensile test data, isoaxial biaxial tensile test data, and uniaxial compression test data. Fatigue crack propagation data include fatigue crack propagation rate at wide tear energy, critical tear energy, and / or fatigue threshold; tensile fatigue life data requires at least three strains and at least seven parallel specimens at each strain.

[0029] Example 2 This embodiment provides a detailed explanation of the fitting of the hyperelastic constitutive model in step 2 and the writing of the UHYPER subroutine.

[0030] Existing technologies rarely apply multiple hyperelastic constitutive models to fatigue life analysis, typically relying on only a few classic models built into finite element software. These models have significant limitations in describing increasingly complex stress-strain curves in rubber. This invention enables the application of a custom hyperelastic constitutive model to fatigue life analysis of rubber materials, greatly expanding the selection range of hyperelastic constitutive models and providing technical support for obtaining high-precision fatigue life prediction results.

[0031] Based on the corresponding mechanical experimental data, the selected hyperelastic constitutive models were fitted with parameters. Specifically, for all hyperelastic constitutive model parameters in Table 1, the least squares method was used for fitting.

[0032] in, Let be the objective function. i =1,2,…,n, where n is the total number of experimental data points. Let i be the nominal stress value obtained from the i-th experimental measurement. For the i-th predicted nominal stress value, The arithmetic mean of all experimental stress values. .

[0033] The stress-strain relationship under uniaxial tensile loading mode follows the formula:

[0034]

[0035] The stress-strain relationship under planar tensile loading mode follows the formula:

[0036]

[0037]

[0038] The stress-strain relationship under isoaxial tensile loading mode follows the formula:

[0039] . In the formula, , , The subscripts 1, 2, and 3 correspond to the three main elongation directions, respectively. The nominal stress under uniaxial tensile loading mode. The nominal stress under planar tensile loading mode, The nominal stress is under the isoaxial tensile loading mode. and These are the first strain invariant and the second strain invariant, respectively. W Let be the strain energy density function. λ Main elongation ratio, W / , W / For the partial derivative of strain energy with respect to invariants.

[0040] When step 1 only provides uniaxial stretching data, for models with many parameters (such as Alexander, Amin, etc.), fix the volume parameters or some higher-order parameters, or use typical values ​​from the literature, and then fit the remaining parameters using the least squares method.

[0041] In this embodiment, the hyperelastic constitutive model is a strain energy density function model based on strain invariants, and its formula is as follows:

[0042] in, It is a skewed hyperelastic term. For volume hyperelasticity, = / , = / To correct for the strain invariant, J represents the volume change J= .

[0043]

[0044]

[0045]

[0046] in, As the first strain invariant, As the second strain invariant, As the third strain invariant, Indicates the material is in The main stretch ratio in the direction.

[0047] The hyperelastic constitutive model in this invention includes 36 models based on strain invariants, as detailed in the table below: Table 1: Models based on strain invariants

[0048] A fitting procedure for the parameters of the hyperelastic constitutive model is established based on the stress-strain relationship formulas corresponding to the loading modes. The fitting accuracy is determined by the coefficient of determination. R 2 express:

[0049] in, Let n be the objective function, and n be the total number of experimental data points. Let i be the nominal stress value obtained from the i-th experimental measurement. For the i-th predicted nominal stress value, The arithmetic mean of all experimental stress values.

[0050] Furthermore, based on the strain energy density function corresponding to different hyperelastic constitutive models, the first derivative, second derivative, and mixed partial derivative of the strain energy density function with respect to the corresponding strain invariants are determined, and the user-defined UHYPER subroutine code is established based on the derivatives.

[0051] The UHYPER subroutine establishes the mapping relationship between the theoretical variables of the hyperelastic constitutive model and the program interface variables. Specifically, it maps the first strain invariant of the hyperelastic constitutive model... Second partial strain invariant and volume change ratio J Map these parameters to the UHYPER interface variables BI1, BI2, and AJ respectively; define the material parameters of each hyperelastic constitutive model accordingly and input the fitting parameters; map the strain energy density function calculated based on the strain invariants and material parameters to U(1); map the first-order partial derivatives of the strain energy density function with respect to each strain invariant to the UI1 array, and map the second-order partial derivatives and mixed partial derivatives of the strain energy density function with respect to each strain invariant to the UI2 array. W / J is mapped to UI3(1), and W / Mapped to UI4(1). Thus, according to the UHYPER data interface format, the theoretical expressions of each hyperelastic constitutive model are converted into user-defined constitutive models that can be called by the fatigue life calculation program.

[0052] The subroutine is written in Fortran90 and needs to include six core modules: (1) Code documentation: containing complete comments, indicating the name of the hyperelastic constitutive model, the time of proposal, author information, and the mathematical expression of the strain energy function. (2) Standardized data interface: establishing a data communication protocol with the Endurica fatigue solver, strictly matching the variable symbols, data types, and memory allocation structure required by the solver. (3) Material parameter module: defining and reading the isotropic hyperelastic material parameters obtained through curve fitting as the input basis for model calculation. (4) Strain energy function construction: writing the partial term and volume term expressions of the strain energy density function according to the selected theoretical model. (5) Derivative calculation core: numerically solving the first, second, and mixed partial derivatives of the strain invariants corresponding to the strain energy function. (6) Program return module: after the integral point calculation is completed, reasonably releasing resources and returning program control to the main solver.

[0053] Example 3 This embodiment provides a detailed description of step 3. Fitting the corresponding fatigue crack propagation model based on the fatigue crack propagation rate includes: performing parameter fitting on a preset fatigue crack propagation model, which includes the Lake-Lindley model and / or the Thomas model, to obtain fatigue crack propagation model parameters corresponding to the rubber material to be analyzed.

[0054] The `ki.json` file, required for the Endurica fatigue life solver, is created. This `ki.json` configuration file follows JSON syntax rules and includes output control information, material constitutive behavior information, crack precursor size information, fatigue crack propagation model information, and loading history information. The output control information is set through the `outputs` field; the material constitutive behavior information is set in the `materials` field, and the hyperelastic constitutive behavior of the material is specified as `uhyper` through the `type` field in the `elasticity` object, enabling the Endurica solver to call the user-defined `UHYPER` subroutine during fatigue life calculation. The crack precursor size information includes the initial crack size `c0` and the final crack size `cf`; the fatigue crack propagation model information is set through the `crack_growth_rate` object, where the `type` field specifies the type of fatigue crack propagation model, and the corresponding model parameters are written according to the selected fatigue crack propagation model. When using the Lake-Lindley model, the model parameters include the power exponent parameter `F0`, the critical crack driving force `Tcritical`, the critical crack propagation rate `rc`, and the crack propagation threshold `threshold`. The loading history information is set through the history field, which includes the loading history type and at least one loading history object. The loading history object further includes loading steps, data blocks, and loading history data. The loading history to be analyzed is written to the .ki.json file through a nested data structure of history.histories[].steps[].blocks[].data, so that the Endurica solver can read it during fatigue life analysis.

[0055] In one implementation, historical loading data is directly written to the `history` field of the `.ki.json` file. In another implementation, the historical loading data is stored in a separate `.ki.db` file, and the filename or storage path of the `.ki.db` file is specified by the `histories` field of the `history` object in the `.ki.json` file. The `.ki.db` file primarily stores the calculation results from finite element analysis software, including Abaqus and / or Ansys. The historical loading configuration information is defined through the `history` field, which includes an `options` subfield and a `histories` subfield. The `options` subfield is used to set the loading cycle and calculation temperature, where the `period` field sets the time period corresponding to one loading cycle, and the `temperature` field sets the temperature corresponding to the fatigue life analysis. The `histories` field specifies an external historical loading data file, and its value is the filename or file path of the `.ki.db` file. The Endurica solver reads the corresponding `.ki.db` file based on the `histories` field to obtain the strain history loading data required for fatigue life calculation. During implementation, the `.db` file can be exported from the mechanical finite element software Abaqus or Ansys.

[0056] Specifically, the fatigue crack propagation model is either the Lake-Lindley model or the Thomas model. The formula for the Lake-Lindley model is as follows:

[0057] In this formula, Defined as a very small crack propagation rate, below which the load will not cause the crack to propagate, but environmental factors such as ozone can cause molecular chains to break, resulting in an extremely small crack propagation rate.

[0058]

[0059]

[0060]

[0061] Thoams model formula:

[0062]

[0063] in, The fatigue threshold, To transform tearing energy, The critical tearing energy, The critical spread rate, F Let A be the power-law exponent and A be the slope of the linear region.

[0064] Furthermore, the ki.json file required for calculations in Endurica software must conform to JSON syntax rules and include the output module, specifying the hyperelastic constitutive behavior of the material as uhyper, setting the fatigue crack propagation model type and parameters, crack precursor body size, temperature-dependent parameters, and compiling strain history loading data and / or specifying it as a .db file.

[0065] Example 4 This embodiment provides a detailed description of step 3. In this step, multiple Fortran source files containing material constitutive relations and corresponding model parameters are obtained. Based on file identifiers, each constitutive model source file is associated with multiple fatigue analysis configuration files corresponding to different load levels. The integrity of the associated files is verified, and subsequent processing is performed only on constitutive models that simultaneously have a preset number of load configuration files. The preset number is at least two, preferably three. The compiler is invoked to compile the constitutive model source files that have passed the integrity verification into a dynamic link library that the fatigue analysis program can call. After successful compilation, configuration files for different load levels are sequentially input into the fatigue analysis program to calculate the fatigue life under the corresponding load conditions. After all load conditions are calculated, a results directory is created according to the constitutive model identifier. The constitutive model files, load configuration files, and fatigue calculation results are archived in the corresponding results directory, and the compilation status, calculation status, and error information of each constitutive model are recorded. By establishing an automatic association between constitutive model source files and multi-load fatigue analysis configuration files, as well as a compilation result reuse mechanism, each constitutive model only needs to be compiled once to continuously complete fatigue life calculations for multiple load levels. This reduces repeated compilation, manual configuration, and calculation result organization operations in the process of multi-constitutive model and multi-load level fatigue analysis, improves the efficiency of batch fatigue life analysis, and reduces the risk of file mismatch caused by manual operation.

[0066] Further, step 4 involves performing multi-load fatigue batch calculations on candidate constitutive models, including obtaining Fortran source files corresponding to multiple user-defined UHYPER constitutive models and multiple fatigue analysis configuration files corresponding to each Fortran source file; extracting constitutive model identifiers from each Fortran source file according to preset file naming rules, and searching for multiple fatigue analysis configuration files with the same constitutive model identifier, thereby establishing a file group correspondence between constitutive model source files and multiple fatigue analysis configuration files with different load levels; performing integrity checks on the established file groups, and determining the file group as a valid file group when the number of fatigue analysis configuration files corresponding to a constitutive model reaches a preset number, otherwise skipping the file group and recording abnormal information; and performing the calculation once on the Fortran source files in each valid file group. The system compiles and generates a user-defined constitutive model compilation file for the fatigue life solver, and verifies the compilation results. When compilation is successful, the fatigue life solver is called sequentially according to a preset order, and multiple fatigue analysis configuration files corresponding to the same constitutive model are submitted to the fatigue life solver to complete fatigue life calculations under different load levels. The calculation execution status of each fatigue analysis configuration file is monitored, and the calculation of the next load level continues if the current load level calculation is successful. If any load level calculation fails, the subsequent calculation of the current constitutive model file group is stopped and the failure information is recorded. After all load level calculations corresponding to the same constitutive model are completed, a result storage directory is established according to the constitutive model identifier, and the corresponding constitutive model source file, fatigue analysis configuration file, compiled file, and fatigue life calculation result file are archived to the result storage directory.

[0067] In this configuration, multiple fatigue analysis configuration files are named using the format "constitutive model identifier + load level identifier". The Fortran source file corresponding to the same constitutive model and multiple fatigue analysis configuration files have the same constitutive model identifier, while different fatigue analysis configuration files are distinguished by different load level identifiers. In one embodiment, each constitutive model corresponds to three fatigue analysis configuration files, which are used for fatigue life calculation under three preset strain levels.

[0068] For the file group that passes integrity verification, the constitutive model source files are compiled to generate a constitutive model library that can be called by the fatigue analysis solver. Specifically, the compiler is invoked to compile the constitutive model source files that pass integrity verification into a dynamic link library that can be called by the fatigue analysis program. The fatigue analysis solver is controlled to reuse the constitutive model library, sequentially read multiple fatigue analysis configuration files in the file group, and obtain fatigue response data under the corresponding load levels; the configuration files of different load levels are sequentially input into the fatigue analysis program to calculate the fatigue life under the corresponding load conditions. After all load conditions are calculated, a results directory is created according to the constitutive model identifier, and the constitutive model files, load configuration files, and fatigue calculation results are archived to the corresponding results directory, and the compilation status, calculation status, and abnormal information of each constitutive model are recorded.

[0069] Example 5 This embodiment provides a detailed explanation of step 5, which evaluates the accuracy of fatigue life range prediction for different hyperelastic constitutive models.

[0070] Due to the discreteness of crack precursor body size, the fatigue life of rubber materials exhibits significant fluctuations. Based on tensile fatigue experimental data of rubber materials and the fatigue life prediction ranges of various hyperelastic constitutive models under different strains, the normalized area method is used to quantitatively evaluate the prediction accuracy of fatigue life ranges for different hyperelastic constitutive models.

[0071] The normalized area method in step 5 evaluates the accuracy of the fatigue life range prediction of the hyperelastic constitutive model. It mainly predicts the tensile fatigue life range of the rubber material at different strain levels based on the range of crack precursor body size, and compares this prediction with tensile fatigue experimental data. The normalized area formula is as follows:

[0072]

[0073]

[0074] in, and These refer to the maximum and minimum experimental fatigue life, respectively. It is the normalized area under a specific strain. It is the average normalized area under multiple strains. It is the standard deviation of the normalized area under multiple strain levels. It is the area enclosed by the line segment formed by the experimental life and the predicted life, and the ideal line (y=x). The smaller the area, the higher the accuracy of the fatigue life range prediction; the smaller the standard deviation, the higher the consistency of the prediction accuracy under multiple strains.

[0075] When inputting the size of the crack precursor, the maximum and minimum values ​​of the crack precursor size distribution are input respectively. From this, the minimum and maximum values ​​of the predicted lifetime are predicted, and the range of the predicted lifetime is determined. , The experimental lifetime range is []. , In the graph of predicted lifetime versus experimental lifetime, the horizontal axis represents experimental lifetime, and the vertical axis represents predicted lifetime. Therefore, the two points on the line segment are A=( , ) = [ , ] and B = ( , ) = [ , These two points form a line segment. Then, define the signed longitudinal deviation of the two endpoints relative to the ideal line y=x: dA= - = - dB= - = - Therefore, the area of ​​a line segment and an ideal line can be divided into two cases: trapezoid and triangle.

[0076] When the two endpoints of the predicted line segment are on the same side of the ideal line, the area is calculated using the trapezoidal method, i.e.: .

[0077] Predicting a line segment's crossing of an ideal line—calculate the area using two triangles. , This is the intersection point of the line segment and the ideal line.

[0078] The average normalized area corresponding to multiple preset strain levels is calculated to obtain the average normalized area evaluation index of the candidate hyperelastic constitutive model. The smaller the average normalized area, the closer the fatigue life range predicted by the candidate hyperelastic constitutive model at multiple strain levels is to the experimental fatigue life range. The normalized area method is used to quantitatively evaluate the prediction accuracy of various hyperelastic constitutive models under tensile fatigue, providing a selection scheme for the rational application of hyperelastic constitutive models in subsequent more complex multiaxial fatigue analyses. Furthermore, the values ​​corresponding to multiple strain levels are calculated.

[0079] Example 6: This embodiment describes a custom hyperelastic constitutive rubber fatigue life prediction system constructed using the aforementioned method. The system includes the following modules: The data acquisition module is used to acquire mechanical experimental data, fatigue crack propagation rate data, and tensile fatigue life data of rubber materials under at least one deformation mode and at least three strain levels.

[0080] The constitutive subroutine module is used to perform parameter fitting on various hyperelastic constitutive models based on strain invariants based on mechanical experimental data. For each constitutive model, the module writes the UHYPER subroutine to establish the mapping between theoretical variables and UHYPER interface variables based on its strain energy density function and the first, second, and mixed partial derivatives with respect to the first strain invariant, the second strain invariant, and the volume ratio.

[0081] The configuration file generation module is used to fit the Lake-Lindley model and / or the Thomas model based on fatigue crack propagation rate data, generating the ki.json file used by Endurica. The ki.json file includes at least output control information, material constitutive information, crack precursor body size information, crack propagation model information, and loading history information. The hyperelastic behavior is specified as uhyper in the material constitutive information.

[0082] The batch processing scheduling module is used to associate the UHYPER subroutine source files of each constitutive model with fatigue analysis configuration files for multiple load levels based on file identifiers, and to perform integrity checks on the associated file groups. Simultaneously, it compiles only source files with a preset number of load configuration files into a reusable dynamic link library for the fatigue solver, and sequentially calls the fatigue solver to calculate fatigue life according to load levels. Specifically, the batch processing scheduling module is configured to perform the following operations: Extract the constitutive model identifier from each UHYPER source file, find multiple load configuration files with the same identifier to form a file group; perform integrity verification on the file group, and only call the compiler to generate dynamic link libraries for file groups that reach the preset number of loads; control the fatigue solver to reuse the same dynamic link library and calculate in load order; the result archiving module writes the results to the directory according to the constitutive model identifier and records the compilation log, calculation log and exception log; the accuracy evaluation module outputs the average normalized area sort and standard deviation sort of each constitutive model.

[0083] The results archiving module is used to create a results directory based on the constitutive model identifier, archive source files, configuration files, compiled files and calculation results, and record compilation status, calculation status and exception information.

[0084] The accuracy assessment module is used to obtain the predicted lifetime range of each constitutive model at each strain level based on the maximum and minimum values ​​of the crack precursor size, and to calculate the original area enclosed by the experimental-prediction line segment and the ideal line y = x in combination with the experimental lifetime range. Calculate the normalized area at each strain level. The average normalized area and standard deviation of multiple strain levels are used to evaluate the accuracy of fatigue life range prediction for different constitutive models.

[0085] Example 7: This embodiment uses a natural rubber / butadiene rubber (30 / 70) nanocomposite material as an example to further illustrate the present invention.

[0086] (1) First, obtain the uniaxial tensile stress-strain curve, fatigue crack propagation data, and tensile fatigue data at three strain levels (95%, 131%, and 152%). In this embodiment, only uniaxial tensile data are used to fit 36 ​​hyperelastic constitutive models. For models with many parameters, the volume parameter DD or K0K0 is taken as a typical value from the literature, and the remaining parameters are fitted by the least squares method. The uniaxial tensile test is conducted at room temperature (23℃) and tensile rate (50 mm / min), with a gauge length of 20 mm. The fatigue crack propagation test measures the crack propagation rate over a wide tear energy range. For the tensile fatigue test, at least 7 parallel specimens are tested at each strain, and the test life range is as follows: Figure 2 As shown.

[0087] (2) Based on the uniaxial tensile stress-strain curve, 36 hyperelastic constitutive models were fitted, and the objective function was the coefficient of determination. :

[0088] Fitting accuracy The sorting is as follows Figure 3 As shown. The specific fitting parameters are the applicant's trade secret and will not be disclosed in the specification; however, those skilled in the art can fit the same parameters without creative effort based on the uniaxial tensile test data obtained in step one and the aforementioned least squares objective function. Subsequently, the UHYPER subroutine code for each hyperelastic constitutive model was written in Fortran90 language format, mainly including comments, data communication protocol with the Endurica fatigue solver, material parameter section, strain energy function expression, derivatives of the strain energy function and strain invariants, and program return section.

[0089] Based on fatigue crack propagation data: fatigue threshold, critical tear energy, and fatigue crack propagation rate at various tear energies, a Lake-Lindley model was fitted. The data and fitting parameters are as follows: Figure 4 As shown. The fatigue threshold is obtained by fitting the Lake-Lindley model using the least squares method. Critical tearing energy Critical spread rate And the power law exponent F. Specific fitted values ​​are trade secrets and will not be disclosed. Those skilled in the art can obtain the same parameters based on the crack propagation data obtained in step one, using the Lake-Lindley model formula and the least squares method.

[0090] Simultaneously, the ki.json file required by the Endurica software needs to be written. The ki.json file needs to specify three parts: the output module, the material data module, and the loading history module. The output module includes: life, crack energy density, strain energy density, engineering strain, engineering stress, and material analysis. The material data module mainly includes a hyperelastic constitutive model and a fatigue crack propagation model. The hyperelastic constitutive model type is specified as uhyper, and the fatigue crack propagation model is Lake-Lindley, with the fitted data parameters input. The loading history module sets the element type to "ne_si" to indicate a uniaxial tensile element, and sets the loading strains consistent with tensile fatigue: 95%, 131%, and 152%. Since it is a simple tensile fatigue, the mechanical aspects can be expressed using a single uniaxial tensile element, without needing analysis in finite element software like Abaqus or Ansys, and therefore no .db file is required. For complex multiaxial load fatigue life analysis, a .db file needs to be exported from the finite element software for input.

[0091] (4) Since this invention includes 36 hyperelastic constitutive models, and each model requires calculation of tensile fatigue life at three strain levels, hundreds of files will be generated. This will lead to problems such as time-consuming manual compilation, mismatch between model program and load file, easy omissions in manual calculation, easy confusion of a large number of files, and difficulty in quickly locating specific models and working conditions in case of calculation failure. Therefore, it is necessary to establish an automated process that links "constitutive model compilation - load configuration matching - Endurica calculation - result archiving" to achieve standardized and automated fatigue calculation of different constitutive models at multiple load levels under unified input naming and calculation rules, so as to provide structurally consistent data for subsequent constitutive model evaluation, fatigue life comparison and optimal model selection. The material stress verification data output by the UHYPER subroutine in the Endurica solver is as follows: Figure 5 As shown, the fatigue life prediction range based on the hyperelastic constitutive model at three strain levels under tensile fatigue is compared with experimental data. Figure 6 .

[0092] (5) By Figure 6As can be seen, the fatigue life prediction range and experimental data are represented as geometric line segments in the graph, and there is an order-of-magnitude difference in life between different peak strain levels. Simply relying on whether the predicted life falls within the 1.5-fold dispersion line is insufficient to quantitatively evaluate the performance of the hyperelastic constitutive model in predicting the fatigue life range. Therefore, the normalized area between the predicted and experimental life line segments and the ideal line (y=x) is calculated to further quantitatively evaluate the performance of each hyperelastic constitutive model. Normalized area method formula:

[0093]

[0094]

[0095] in, and These refer to the maximum and minimum experimental fatigue life, respectively. It is the normalized area under a specific strain. It is the average normalized area under multiple strains. It is the standard deviation of the normalized area under multiple strain levels. This is the area enclosed by the line segment formed by the experimental life and the predicted life, and the ideal line (y=x). A smaller area indicates higher accuracy in predicting the fatigue life range, and a smaller standard deviation indicates higher consistency in prediction accuracy across multiple strains. The average normalized area of ​​the 36 hyperelastic constitutive models is shown below. Figure 7 As shown. Specific values ​​are trade secrets and will not be disclosed. The hyperelastic constitutive model with a lower average normalized area can be used for more complex multiaxial load analysis in the future.

[0096] In summary, this invention can effectively implement various custom hyperelastic constitutive models for fatigue life analysis and evaluation.

[0097] The above technical solution is only one embodiment of the present invention. For those skilled in the art, based on the principles disclosed in the present invention, it is easy to make various types of improvements or modifications, and not limited to the technical solutions described in the specific embodiments of the present invention. Therefore, the foregoing description is only a preferred option and is not restrictive.

Claims

1. A method for predicting fatigue life using a custom rubber constitutive model, characterized in that, The method includes: Step 1: Obtain mechanical experimental data, fatigue crack propagation rate and tensile fatigue life data of rubber materials under at least one deformation mode; Step 2: Based on the mechanical experimental data, perform parameter fitting on the selected hyperelastic constitutive models to obtain the hyperelastic constitutive model parameters corresponding to the rubber material, and write UHYPER subroutine code according to the strain energy density function of the hyperelastic constitutive model and its first derivative, second derivative and mixed partial derivative with respect to strain invariants. Step 3: Fit a fatigue crack propagation model based on the fatigue crack propagation rate, and compile the material model data into the configuration file required by the Endurica solver; Step 4: Perform batch processing of fatigue simulations for multiple hyperelastic constitutive models and multiple load levels, including: obtaining multiple constitutive model source files and their corresponding fatigue analysis configuration files for multiple load levels; establishing the association between the source files and configuration files and performing integrity verification; compiling the constitutive model source files that pass the verification; calculating the fatigue life under each load in sequence and archiving the calculation results; Step 5: Based on the tensile fatigue test data of rubber materials and the fatigue life prediction range of each hyperelastic constitutive model at various strain levels, the normalized area method is used to quantitatively evaluate the fatigue life range prediction accuracy of different hyperelastic constitutive models.

2. The fatigue life prediction method for a custom rubber constitutive model according to claim 1, characterized in that, In step 2, the parameter fitting uses the least squares method, and the fitting accuracy is determined by the coefficient of determination. The formula is as follows: in, For the i-th experimental stress value, For the i-th predicted nominal stress value, denoted as the average experimental stress, and n as the total number of data points.

3. The fatigue life prediction method for a custom rubber constitutive model according to claim 1, characterized in that, The configuration file in step 3 includes: Output control information, used to specify the output items required for fatigue life calculation; Material constitutive information, used to specify the hyperelastic constitutive behavior of the material as UHYPER; Crack precursor size information, including initial crack size and terminal crack size; Fatigue crack propagation model information, used to specify the type and parameters of the fatigue crack propagation model; The loading history data file is used to provide strain history loading data.

4. The fatigue life prediction method for a custom rubber constitutive model according to claim 1, characterized in that, In step 4, the integrity verification operation includes: determining whether the number of fatigue analysis configuration files corresponding to each constitutive model has reached a preset number; performing subsequent processing only on constitutive models that have reached the preset number, and skipping and recording abnormal information for constitutive models that have not reached the preset number. The fatigue analysis configuration file adopts the naming method of constitutive model identifier + load level identifier; the constitutive model source file corresponding to the same constitutive model and multiple fatigue analysis configuration files have the same constitutive model identifier, and different fatigue analysis configuration files are distinguished by different load level identifiers.

5. The fatigue life prediction method for a custom rubber constitutive model according to claim 4, characterized in that, The preset quantity is 3, which corresponds to the fatigue life calculation under three preset strain levels.

6. The fatigue life prediction method for a custom rubber constitutive model according to claim 1, characterized in that, The operation of compiling the source files of the constitutive model that have passed the verification in step 4 includes: calling the compiler to compile the source files of the constitutive model into a dynamic link library that can be called by the fatigue analysis program; and, each source file that has passed the verification is compiled only once, so that it can be reused for multiple load level calculations corresponding to the same constitutive model. The operation of sequentially calculating fatigue life under each load includes: detecting the calculation execution status of each fatigue analysis configuration file; continuing to execute the calculation of the next load level when the current load level calculation is successful; stopping the subsequent calculation of the current constitutive model file group and recording the failure information when any load level calculation fails. The archiving calculation operation includes: creating a result directory according to the constitutive model identifier, archiving the constitutive model source file, fatigue analysis configuration file, compiled generated file and fatigue life calculation result file to the result directory, and recording abnormal information, which includes at least one of compilation failure information, calculation failure information and configuration file missing information.

7. The fatigue life prediction method for a custom rubber constitutive model according to claim 1, characterized in that, Step 5, which uses the normalized area method to quantitatively evaluate the fatigue life range prediction accuracy of different hyperelastic constitutive models, includes the following steps: For each strain level, the predicted fatigue life range and the experimental fatigue life range at that strain level are determined respectively. Using experimental fatigue life as the abscissa and predicted fatigue life as the ordinate, the area enclosed between the line segment representing the range of predicted fatigue life and the range of experimental fatigue life and the ideal line y=x is taken as the original area under this strain level. The original area is normalized to obtain the normalized area at this strain level; Evaluation indices are calculated based on normalized areas at multiple strain levels. These indices are used to characterize the accuracy of fatigue life range prediction for the hyperelastic constitutive model.

8. The fatigue life prediction method for a custom rubber constitutive model according to claim 7, characterized in that, The original area is calculated in the following way: Define the signed longitudinal deviation dA of the two endpoints of a line segment relative to the ideal line y=x. - dB= - When dA•dB≥0, the original area is calculated as a trapezoid; when dA•dB<0, the original area is calculated as two triangles. The normalized area is calculated using the following formula: in, Normalized area under specific strain and These represent the maximum and minimum values ​​of the experimental fatigue life, respectively. The original area is denoted as .

9. A custom-designed fatigue life prediction system for hyperelastic constitutive rubber, characterized in that, The system includes: The data acquisition module is used to acquire mechanical experimental data, fatigue crack propagation rate data, and tensile fatigue life data of rubber materials under at least one deformation mode and at least three strain levels. The constitutive subroutine module is used to perform parameter fitting on various hyperelastic constitutive models based on strain invariants based on the mechanical experimental data, and to write UHYPER subroutines for each constitutive model according to its strain energy density function and the first, second, and mixed partial derivatives with respect to the first strain invariant, the second strain invariant, and the volume ratio, and to establish the mapping between theoretical variables and UHYPER interface variables. The configuration file generation module is used to fit the Lake-Lindley model and the Thomas model based on the fatigue crack propagation rate data and generate the configuration file used by the Endurica solver. The batch processing scheduling module is used to associate each constitutive model source file with fatigue analysis configuration files for multiple load levels according to the file identifier, and to perform integrity verification on the associated file group; it compiles the source files with a preset number of load configuration files into a reusable dynamic link library for the fatigue solver, and sequentially calls the fatigue solver to calculate fatigue life according to the load level. The results archiving module is used to create a results directory according to the constitutive model identifier, archive source files, configuration files, compiled files and calculation results, and record compilation status, calculation status and exception information; The accuracy assessment module is used to obtain the predicted lifetime range of each constitutive model at various strain levels based on the maximum and minimum values ​​of the crack precursor size. Combined with the experimental lifetime range, it calculates the original area enclosed by the experimental-prediction line segment and the ideal line y=x. The normalized area of ​​each strain level and the average normalized area and standard deviation of multiple strain levels are calculated to evaluate the accuracy of fatigue life range prediction for different constitutive models.

10. The custom rubber constitutive model fatigue life prediction system according to claim 9, characterized in that, In the batch processing scheduling module, the integrity verification operation includes: determining whether the number of fatigue analysis configuration files corresponding to each constitutive model has reached a preset number; only performing compilation and calculation on constitutive models that have reached the preset number, and skipping constitutive models that have not reached the preset number and recording abnormal information.