Creep-fatigue life prediction method and readable storage medium

Abstract

The invention relates to a creep-fatigue life prediction method and a readable storage medium. The creep-fatigue life prediction method comprises the following steps: S1, acquiring a creep-fatigue small sample data set of a test material; s2, constructing a life prediction model by using the small sample data set and verifying the life prediction model; s3, obtaining a large sample data set through a finite element data simulation method; s4, machine learning is adopted, a large sample data set is used for training, and a creep-fatigue life prediction platform is obtained; and S5, inputting the data of the to-be-predicted test material into the creep-fatigue life prediction platform to obtain a prediction value and evaluate the prediction performance. The method is based on small sample data construction, and has the advantages of high prediction speed, high prediction precision and wide prediction range.

Description

Technical Field

[0001] The technical field of the present invention relates to a creep-fatigue life prediction method and a readable storage medium. Background Art

[0002] High-temperature components such as aircraft engine turbine disks, steam turbine rotors, and power plant boilers are not only subjected to constant temperature and stress loads, but also usually to alternating loads caused by start-stop and temperature fluctuations. Therefore, they are often accompanied by obvious creep-fatigue load interactions during actual service, leading to damage and failure. The life of materials under creep-fatigue load interactions is much lower than that under single fatigue loads or single creep loads. Therefore, when creep and fatigue affect at the same time, it is necessary to consider not only their individual effects, but also the effects of their interactions, in order to more accurately predict component life.

[0003] At present, most of the research on the creep-fatigue behavior of high-temperature materials is carried out through experiments, theories and numerical simulations. However, the cost of creep-fatigue experiments is high, and only small sample data can be obtained. The traditional creep-fatigue life prediction theoretical model requires a large amount of test data to fit the parameters, can only be predicted under a single working condition, and it is very time-consuming to perform a complete creep-fatigue numerical simulation. Machine learning, as a data-driven solution, can take into account a variety of complex factors and reduce computing costs. It has been applied to various fields such as constitutive modeling and life prediction of materials. Summary of the invention

[0004] The purpose of the present invention is to provide a creep-fatigue life prediction method.

[0005] Another object of the present invention is to provide a readable storage medium.

[0006] According to one aspect of the present invention, a creep-fatigue life prediction method includes: S1. obtaining a small sample data set of creep-fatigue of a test material; S2. using the small sample data set to construct a life prediction model and verify it; S3. obtaining a large sample data set through a finite element data simulation method; S4. using machine learning to train with the large sample data set to obtain a creep-fatigue life prediction platform; S5. inputting the data of the material to be predicted into the creep-fatigue life prediction platform to obtain a predicted value and evaluate the prediction performance.

[0007] The technical solution of the present application obtains a small sample data set of creep-fatigue of the uniaxial and multi-axial stress states of the test material, constructs a life prediction model based on continuous damage mechanics and verifies the small sample data set, and then obtains a large amount of multi-condition data through finite element numerical simulation, generates creep-fatigue data of a sufficient scale to train and test the machine learning model, and obtains an accurate and generalizable data-driven creep-fatigue life prediction platform, which has low prediction calculation cost, is built based on small sample data, and has the characteristics of fast prediction speed, high prediction accuracy and wide prediction range.

[0008] In one or more embodiments of the creep-fatigue life prediction method, in step S1, the test material includes uniaxial and multiaxial specimens.

[0009] In one or more embodiments of the creep-fatigue life prediction method, in step S1, the small sample data set includes the stress concentration factor K of the test material. t , creep-fatigue load condition and creep-fatigue life, wherein the creep-fatigue load condition includes the strain ratio R ε , total strain range Δε t and tensile holding time t h .

[0010] In one or more embodiments of the creep-fatigue life prediction method, in step S2, the life prediction model includes a fatigue damage model and a creep damage model;

[0011] The fatigue damage model is:

[0012]

[0013] σ 0 is the maximum stress for each cycle, Δε p is the plastic strain range of each cycle, a and b are temperature-related material parameters;

[0014] The creep damage model is:

[0015]

[0016]

[0017]

[0018] i is the number of incremental steps in the numerical simulation, is the creep strain energy density rate, w f is the failure strain energy density, w f,crit is the critical strain energy density, is the average stress within a cycle, is the equivalent peak tensile stress, ε c is the creep strain, τ is the time increment required for each analysis step, m c and n c are material and temperature related model parameters, respectively.

[0019] In one or more embodiments of the creep-fatigue life prediction method, in step S2, the verification includes calculating the numerical value of the life prediction model by a finite element numerical simulation method to obtain the creep-fatigue life of the test material, and comparing it with the test results.

[0020] In one or more embodiments of the creep-fatigue life prediction method, in S3, the large sample data set includes input variables and output variables, and the input variables include the stress concentration factor K t , strain ratio R ε , total strain range Δε t and tensile holding time t h , the output variables include creep-fatigue life N.

[0021] In one or more embodiments of the creep-fatigue life prediction method, in S4, the machine learning includes a support vector machine model, the large sample data set is divided into a training group and a test group, the hyperparameters of the support vector machine model are optimized by cross-validation, and the creep-fatigue life prediction platform is the support vector machine model after the hyperparameters are optimized.

[0022] In one or more embodiments of the creep-fatigue life prediction method, in S5, the data of the material to be predicted includes the input variables, and the data of the material to be predicted is normalized and then input into the creep-fatigue life prediction platform.

[0023] In one or more embodiments of the creep-fatigue life prediction method, in S5, the prediction performance is evaluated by determining the coefficient R 2 and mean absolute percentage error MAPE, the coefficient of determination R 2 And the expression of the mean absolute percentage error MAPE is as follows:

[0024]

[0025]

[0026] is the predicted creep-fatigue life, y i is the corresponding test result, y meanRepresents the average value of creep-fatigue test life.

[0027] According to another aspect of the present invention, a readable storage medium stores a computer program, which is executed by a processor to implement any of the creep-fatigue life prediction methods described above. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] The above and other features, properties and advantages of the present invention will become more apparent through the following description in conjunction with the accompanying drawings and embodiments. In the accompanying drawings, the same reference numerals always represent the same features. It should be noted that these drawings are only examples and are not drawn according to the conditions of equal scale, and should not be used as a limitation on the actual scope of protection required by the present invention, wherein:

[0029] Figure 1 The figure is a flow chart of a creep-fatigue life prediction method according to an embodiment.

[0030] Figure 2 This is a comparison diagram of creep-fatigue life predicted by a creep-fatigue life prediction method and test life in one embodiment.

[0031] Figure 3 The present invention is a schematic diagram of the influence of creep-fatigue load conditions on predicted life obtained by using a creep-fatigue life prediction method according to an embodiment.

[0032] Reference numerals:

[0033] 100-Creep-Fatigue Life Prediction Method. DETAILED DESCRIPTION

[0034] Reference will now be made in detail to various embodiments of the present invention, examples of which are shown in the accompanying drawings and described below. Although the present invention will be described in conjunction with the exemplary embodiments, it should be appreciated that this specification is not intended to limit the present invention to those exemplary embodiments. On the contrary, the present invention is intended to cover not only these exemplary embodiments, but also various alternative forms, modifications, equivalent forms and other embodiments that may be included within the spirit and scope of the present invention as defined by the appended claims.

[0035] At the same time, the present application uses specific words to describe the embodiments of the present application. For example, "one embodiment" and / or "an embodiment" means a certain feature, structure or characteristic related to at least one embodiment of the present application. Therefore, it should be emphasized and noted that "one embodiment" or "an embodiment" mentioned twice or more in different positions in this specification does not necessarily refer to the same embodiment. In addition, certain features, structures or characteristics in one or more embodiments of the present application can be appropriately combined.

[0036] Flowcharts are used in this application to illustrate the operations performed by the system according to the embodiments of the application. It should be understood that the preceding or following operations are not necessarily performed precisely in order. Other operations may also be added to these processes, or one or more operations may be removed from these processes.

[0037] Although machine learning can solve the problem of high computational cost and long time consumption of traditional methods for creep-fatigue life prediction of high-temperature components, the life prediction method based on machine learning requires a large amount of reliable data for training. The experimental cost of creep-fatigue is high, and only small sample data can be obtained. The data in the literature is scattered, so it is impossible to obtain an accurate and reliable prediction method. At the same time, for some areas with discontinuous geometric structures in actual high-temperature components, it is the critical position where creep-fatigue cracks are most likely to initiate. Traditional methods cannot quickly predict creep-fatigue life under multi-axial stress states.

[0038] Based on the above considerations, the inventors conducted in-depth research and designed a creep-fatigue life prediction method. By obtaining a small sample data set of creep-fatigue in the uniaxial and multi-axial stress states of the test material, a life prediction model based on continuous damage mechanics was constructed and the small sample data set was verified. A large amount of multi-condition data was obtained through finite element numerical simulation, and creep-fatigue data of sufficient scale was generated to train and test the machine learning model. An accurate and generalized data-driven creep-fatigue life prediction platform was obtained. The prediction calculation cost is low, and it is built based on small sample data, with the characteristics of fast prediction speed, high prediction accuracy and wide prediction range.

[0039] refer to Figure 1 As shown, in some embodiments, the specific steps of the creep-fatigue life prediction method may include:

[0040] S1. Obtain a small sample data set of creep-fatigue of the test material.

[0041] Specifically, the test materials include uniaxial and multiaxial specimens to meet the creep-fatigue life prediction of multiaxial stress states in areas with geometric discontinuities in actual high-temperature components, with a wide prediction range. The small sample data set includes the stress concentration factor K of the test material t , creep-fatigue load conditions and creep-fatigue life. Creep-fatigue load conditions include strain ratio R ε , total strain range Δε t and tensile holding time t h In one embodiment, the small sample data set includes twenty-six groups of data.

[0042] S2. Use small sample data sets to build and validate lifespan prediction models;

[0043] Specifically, the life prediction model is based on continuum damage mechanics, and the life prediction model includes a fatigue damage model and a creep damage model.

[0044] The fatigue damage model is a damage model of net tensile hysteresis energy:

[0045]

[0046] σ 0 is the maximum stress for each cycle, Δε p is the plastic strain range for each cycle, and a and b are temperature-related material parameters.

[0047] The creep damage model is a strain energy density depletion model:

[0048]

[0049]

[0050]

[0051] i is the number of incremental steps in the numerical simulation, is the creep strain energy density rate, w f is the failure strain energy density, w f,crit is the critical strain energy density, is the average stress within a cycle, is the equivalent peak tensile stress, ε c is the creep strain, τ is the time increment required for each analysis step, m c and n c are material and temperature related model parameters, respectively.

[0052] Calibrate the parameters of the life prediction model and determine a, b, and m through experiments. c and n c After the parameters are equal, the creep damage d of each node of the analyzed component in the i-th cycle is obtained by finite element calculation. c,macro , fatigue damage f , thus obtaining the cumulative values ​​of the two types of damage in the first i cycles. When the linear superposition value of the accumulated fatigue damage and creep damage at the critical point of the analyzed component (such as the node near the root of the hole) reaches 1, the cycle at this time is determined as the creep-fatigue crack initiation life, and the material fails; if the superposition value does not reach 1, the finite element analysis and damage calculation of the (i+1)th cycle are performed. After the creep-fatigue life of the test material is obtained by the finite element data simulation method, it is compared with the test results (i.e. the creep-fatigue life in the small sample data set) to verify the accuracy of the life prediction model.

[0053] S3. Obtain a large sample data set through finite element data simulation method.

[0054] Specifically, a large amount of multi-condition uniaxial and multiaxial creep-fatigue data is obtained through finite element data simulation to generate a sufficient large sample data set for machine learning. The large sample data set includes input variables and output variables. The input variables include the stress concentration factor K t , strain ratio R ε , total strain range Δε t and tensile holding time t h , the output variables include creep-fatigue life N.

[0055] S4. Use machine learning and large sample data sets for training to obtain a creep-fatigue life prediction platform.

[0056] Specifically, machine learning includes a support vector machine model. The large sample data set is divided into a training group and a test group. The hyperparameters of the support vector machine model are optimized by cross-validation. The creep-fatigue life prediction platform is the support vector machine model after the hyperparameters are optimized. The hyperparameters that need to be optimized for the support vector machine model include the penalty parameter c and the parameter g of the kernel function. The data set is divided into several subsets, one of which is used as the test group, and the remaining subsets are used as the training group. For each hyperparameter combination (parameter c and parameter g), the training group can be used to train the model, and the test group can be used to evaluate the performance of the model. This process is repeated many times, each time using a different subset as the test group, and finally an average performance evaluation result is obtained. The parameters c and g with the best performance are selected as the hyperparameters of the final support vector machine model, and then the final data-driven creep-fatigue life prediction platform is obtained.

[0057] In one embodiment, the large sample data set includes two hundred sets of data, 80% of which are used for training and 20% of which are used for validation.

[0058] S5. Input the data of the material to be predicted into the creep-fatigue life prediction platform, obtain the predicted value and evaluate the prediction performance.

[0059] Specifically, the data of the material to be predicted include the stress concentration factor K t , strain ratio R ε , total strain range Δε t and tensile holding time t h , the data of the material to be predicted is normalized and then input into the creep-fatigue life prediction platform to obtain the creep-fatigue life prediction value of the material to be predicted.

[0060] The prediction performance was evaluated by the coefficient of determination R 2 and mean absolute percentage error MAPE, coefficient of determination R 2The expression of mean absolute percentage error MAPE is as follows:

[0061]

[0062]

[0063] is the predicted creep-fatigue life, y i is the corresponding test result, y mean Represents the average value of creep-fatigue test life.

[0064] R 2 The value of is between 0 and 1. The closer its value is to 1, the better the prediction performance. The value of MAPE is inversely proportional to the prediction accuracy. The smaller the MAPE value, the better the prediction performance. Figure 2 The figure shows the comparison between the creep-fatigue life prediction value and the test life. The prediction performance R of the support vector machine model is obtained by calculation. 2 The value of MAPE is 0.892 and 0.268, and the data points fall within the 2-fold error band. The creep-fatigue life prediction platform has good prediction performance.

[0065] In one embodiment, if Figure 3 As shown, the creep-fatigue life prediction platform of this application is used, and the stress concentration factor K is input t , strain ratio R ε , total strain range Δε t and tensile holding time t h , the output creep-fatigue life N is a schematic diagram of the influence of the operating parameters on the predicted life. It can be seen that with the increase of the holding time, the creep-fatigue life of the material gradually decreases.

[0066] Although the above method is illustrated and described as a series of actions for simplicity of explanation, it should be understood and appreciated that these steps are not limited by the order of the actions, because according to one or more embodiments, some actions may occur in a different order and / or concurrently with other actions from those illustrated and described herein or not illustrated and described herein but understandable to those skilled in the art.

[0067] The present invention also relates to a readable storage medium having a computer program stored thereon. When the program is executed by a processor, the steps executed by the program in the creep-fatigue life prediction method described in the above embodiment can be implemented.

[0068] Although the present invention is disclosed as above with preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art may make possible changes and modifications without departing from the spirit and scope of the present invention. Therefore, any modification, equivalent change and modification made to the above embodiments according to the technical essence of the present invention without departing from the content of the technical solution of the present invention shall fall within the protection scope defined by the claims of the present invention.

Claims

1. A creep-fatigue life prediction method, It is characterized in that include: S1. Obtain a small sample data set of creep-fatigue of the test material; S2. constructing and verifying a lifespan prediction model using the small sample data set; S3. Obtain a large sample data set through finite element data simulation method; S4. Using machine learning to train with the large sample data set to obtain a creep-fatigue life prediction platform; S5. Input the data of the material to be predicted into the creep-fatigue life prediction platform, obtain the predicted value and evaluate the prediction performance.

2. The creep-fatigue life prediction method according to claim 1, It is characterized in that In the step S1, the test materials include uniaxial and multiaxial specimens.

3. The creep-fatigue life prediction method according to claim 2, It is characterized in that In step S1, the small sample data set includes the stress concentration factor K of the test material. t , creep-fatigue load condition and creep-fatigue life, wherein the creep-fatigue load condition includes the strain ratio R ε , total strain range Δε t and tensile holding time t h .

4. The creep-fatigue life prediction method according to claim 1, It is characterized in that In the step S2, the life prediction model includes a fatigue damage model and a creep damage model; The fatigue damage model is: σ 0 is the maximum stress for each cycle, Δε p is the plastic strain range of each cycle, a and b are temperature-related material parameters; The creep damage model is: i is the number of incremental steps in the numerical simulation, is the creep strain energy density rate, w f is the failure strain energy density, w f,crit is the critical strain energy density, is the average stress within a cycle, is the equivalent peak tensile stress, ε c is the creep strain, τ is the time increment required for each analysis step, m c and n c are material and temperature related model parameters, respectively.

5. The creep-fatigue life prediction method according to claim 1, It is characterized in that In the step S2, the verification includes calculating the value of the life prediction model by a finite element numerical simulation method to obtain the creep-fatigue life of the test material, and comparing the calculated value with the test result.

6. The creep-fatigue life prediction method according to claim 1, It is characterized in that In S3, the large sample data set includes input variables and output variables, and the input variables include stress concentration factor K t , strain ratio R ε , total strain range Δε t and tensile holding time t h , the output variables include creep-fatigue life N.

7. The creep-fatigue life prediction method according to claim 1, It is characterized in that In S4, the machine learning includes a support vector machine model, the large sample data set is divided into a training group and a test group, the hyperparameters of the support vector machine model are optimized by cross-validation method, and the creep-fatigue life prediction platform is the support vector machine model after the hyperparameters are optimized.

8. The creep-fatigue life prediction method according to claim 6, It is characterized in that In S5, the data of the material to be predicted includes the input variables, and the data of the material to be predicted is normalized and then input into the creep-fatigue life prediction platform.

9. The creep-fatigue life prediction method according to claim 1, It is characterized in that In S5, the prediction performance is evaluated by the coefficient of determination R 2 and mean absolute percentage error MAPE, the coefficient of determination R 2 And the expression of the mean absolute percentage error MAPE is as follows: is the predicted creep-fatigue life, y i is the corresponding test result, y mean Represents the average value of creep-fatigue test life.

10. A readable storage medium having a computer program stored thereon, It is characterized in that The program is executed by a processor to implement the creep-fatigue life prediction method as described in any one of claims 1 to 9.

Images