Corrosion fatigue prediction model training method, corrosion fatigue prediction model training system, corrosion fatigue prediction method and corrosion fatigue prediction equipment

Abstract

The invention provides a training method, system, prediction method and equipment of a corrosion fatigue prediction model, and the method comprises the steps: obtaining a plurality of initial sample data of aluminum alloy corrosion fatigue, and building an initial data set, the initial sample data comprising a plurality of feature values and corresponding true values of corrosion fatigue crack propagation rates; inputting the partial characteristic value of each piece of initial sample data into a preset parameterized physical model to generate a corresponding predicted value of the corrosion fatigue crack growth rate; constructing a sample data set based on all characteristic values of each piece of initial sample data in the initial data set and a ratio between a true value and a predicted value of the corresponding corrosion fatigue crack growth rate; and training the initial prediction model by using the sample data set to generate an aluminum alloy corrosion fatigue prediction model. The aluminum alloy corrosion fatigue failure behavior prediction precision is improved, and the service life of the aluminum alloy can be accurately predicted.

Description

Technical Field

[0001] The present invention relates to the technical field of material corrosion fatigue life prediction, and in particular to a training method, system, prediction method and equipment for a corrosion fatigue prediction model. Background Art

[0002] Aluminum alloy materials are widely used in aircraft structural materials due to their advantages such as low density, high strength and low cost. However, the high temperature, high humidity and high salt spray climate environment will cause corrosion damage to the surface of aircraft structures serving in coastal areas. Under the action of cyclic loads, these corrosion damaged parts will accelerate the expansion and extension of fatigue cracks and shorten the fatigue life of aircraft structures. Therefore, studying the degradation law of fatigue life of aluminum alloy materials under corrosion damage conditions is of great significance to maintaining the integrity and reliability of aircraft structures.

[0003] In the existing technology, the study of corrosion fatigue is mostly based on the perspective of fracture mechanics, and the fatigue life is predicted by studying the law of fatigue crack propagation of materials under corrosion damage conditions. However, since corrosion fatigue behavior involves multi-factor coupling, the existing models have problems such as difficulty in describing influencing factors, lack of universality and low accuracy. Overestimation or underestimation of material and structural failure behavior data will lead to huge economic losses or catastrophic safety problems. Therefore, there is room for improvement. Summary of the invention

[0004] In view of the shortcomings of the prior art mentioned above, the purpose of the present invention is to provide a training method, system, prediction method and equipment for a corrosion fatigue prediction model, so as to solve the technical problems in the prior art that the corrosion fatigue prediction model of aluminum alloy corrosion has difficulties in describing the influencing factors, lacks universality and has low accuracy.

[0005] To achieve the above-mentioned object and other related objects, the present invention provides a training method for a corrosion fatigue prediction model, comprising:

[0006] Acquire multiple initial sample data of aluminum alloy corrosion fatigue and establish an initial data set, wherein the initial sample data includes multiple characteristic values ​​and corresponding true values ​​of corrosion fatigue crack growth rates;

[0007] Inputting partial characteristic values ​​of each initial sample data into a preset parameterized physical model to generate a corresponding prediction value of the corrosion fatigue crack growth rate;

[0008] Constructing a sample data set based on all characteristic values ​​of each initial sample data in the initial data set and the ratio between the corresponding true value and the predicted value of the corrosion fatigue crack growth rate;

[0009] The sample data set is used to train the initial prediction model to generate an aluminum alloy corrosion fatigue prediction model.

[0010] In one embodiment of the present invention, the step of acquiring a plurality of initial sample data of aluminum alloy corrosion fatigue and establishing an initial data set, wherein the initial sample data includes a plurality of characteristic values ​​and corresponding true values ​​of corrosion fatigue crack growth rates, comprises:

[0011] Multiple initial sample data obtained for aluminum alloy corrosion fatigue testing;

[0012] Randomly generate multiple sets of eigenvalues, input the multiple sets of eigenvalues ​​into a preset sample expansion model, generate corresponding true values, and form multiple expanded initial sample data based on the multiple sets of eigenvalues ​​and the corresponding true values;

[0013] An initial data set is established based on the expanded initial sample data and the acquired multiple initial sample data.

[0014] In one embodiment of the present invention, the parameterized physical model is generated according to the following steps:

[0015] The yield strength in the material parameters of the aluminum alloy, the temperature in the environmental parameters, the range of the stress intensity factor in the mechanical parameters, the loading frequency in the mechanical parameters, and the stress ratio in the mechanical parameters are obtained as modeling features;

[0016] Based on a plurality of the modeling features, a parameterized physical model is constructed.

[0017] In one embodiment of the present invention, the parameterized physical model f KN (x i ) satisfies the following formula:

[0018] f KN (x i ) = f YS ×f T ×f N

[0019] Among them, x i represents the initial sample containing multiple eigenvalues, f YS represents the parameterized model of yield strength in terms of material parameters, f T represents the parameterized model about the temperature in the environmental parameters, f N Represents a parameterized model with respect to the range of stress intensity factors, loading frequency, and stress ratios in the mechanical parameters.

[0020] In one embodiment of the present invention, the step of using the sample data set to train the initial prediction model to generate an aluminum alloy corrosion fatigue prediction model includes:

[0021] Using the XGBoost model as an initial prediction model and setting hyperparameters of the initial prediction model;

[0022] The sample data set is input into the initial prediction model for training, and hyperparameter optimization is performed to generate an aluminum alloy corrosion fatigue prediction model.

[0023] In one embodiment of the present invention, the step of inputting the sample data set into the initial prediction model for training and performing hyperparameter optimization to generate an aluminum alloy corrosion fatigue prediction model includes:

[0024] Dividing the sample data set into a training set and a test set, and using the training set to train the initial prediction model;

[0025] Using the test set to evaluate the trained initial prediction model, and generating a correlation coefficient of the initial prediction model;

[0026] When the correlation coefficient is lower than a preset threshold, adjusting the hyperparameters of the initial prediction model, retraining the adjusted initial prediction model using the training set, and repeatedly evaluating using the test set until the correlation coefficient reaches the preset threshold;

[0027] When the correlation coefficient reaches a preset threshold, the trained initial prediction model is set as the aluminum alloy corrosion fatigue prediction model.

[0028] The present invention also provides a corrosion fatigue prediction model training system, comprising:

[0029] A data acquisition module, used to acquire a plurality of initial sample data of aluminum alloy corrosion fatigue and establish an initial data set, wherein the initial sample data includes a plurality of characteristic values ​​and corresponding true values ​​of corrosion fatigue crack growth rates;

[0030] An initial prediction module is used to input part of the characteristic values ​​of each initial sample data into a preset parameterized physical model to generate a corresponding prediction value of the corrosion fatigue crack growth rate;

[0031] A sample construction module, used to construct a sample data set based on all characteristic values ​​of each initial sample data in the initial data set and the ratio between the actual value and the predicted value of the corresponding corrosion fatigue crack growth rate;

[0032] The model training module is used to train the initial prediction model using the sample data set to generate an aluminum alloy corrosion fatigue prediction model.

[0033] The present invention also provides a method for predicting corrosion fatigue of aluminum alloy, comprising:

[0034] Acquire multiple characteristic values ​​of the aluminum alloy material to be predicted;

[0035] Inputting a plurality of characteristic values ​​into a prediction model for prediction to generate corresponding prediction ratios, wherein the prediction model is any one of the above-mentioned aluminum alloy corrosion fatigue prediction models;

[0036] Some of the multiple eigenvalues ​​are input into a preset parameterized physical model to generate corresponding prediction values, and based on the prediction ratio and the prediction value, the corrosion fatigue crack growth rate of the aluminum alloy is calculated and generated.

[0037] In one embodiment of the present invention, the aluminum alloy corrosion fatigue crack growth rate Satisfies the following formula:

[0038]

[0039] Among them, x i represents multiple eigenvalues, f KN (x i ) represents the predicted value of the parameterized physical model output, C i represents the predicted ratio.

[0040] The present invention also provides an electronic device, comprising:

[0041] one or more processors;

[0042] A storage device for storing one or more programs, which, when executed by the one or more processors, enables the electronic device to implement the training method of the corrosion fatigue prediction model as described in any one of the above or the prediction method of aluminum alloy corrosion fatigue.

[0043] As described above, the training method, system, prediction method and device of a corrosion fatigue prediction model of the present invention have the following beneficial effects: the present invention can solve the technical problems of the difficulty in describing the influencing factors, the lack of universality and the low accuracy of the corrosion fatigue prediction model of aluminum alloy in the prior art. The present invention improves the prediction accuracy of the corrosion fatigue failure behavior of aluminum alloy and can accurately predict its service life. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 A schematic flow chart of a method for training a corrosion fatigue prediction model provided by an embodiment of the present invention;

[0045] Figure 2 A schematic diagram showing the model fitting effect in one embodiment of the present invention;

[0046] Figure 3A schematic diagram showing the results of a model extension test in one embodiment of the present invention;

[0047] Figure 4 Shown is a structural block diagram of a training system for a corrosion fatigue prediction model provided by an embodiment of the present invention;

[0048] Figure 5 Shown is a structural schematic diagram of an electronic device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0049] The following describes the embodiments of the present invention by specific examples, and those skilled in the art can easily understand other advantages and effects of the present invention from the contents disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and the details in this specification can also be modified or changed in various ways based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that the following embodiments and features in the embodiments can be combined with each other without conflict.

[0050] It should be noted that the illustrations provided in the following embodiments are only schematic illustrations of the basic concept of the present invention, and thus the drawings only show components related to the present invention rather than being drawn according to the number, shape and size of components in actual implementation. In actual implementation, the type, quantity and proportion of each component may be changed arbitrarily, and the component layout may also be more complicated.

[0051] In the following description, numerous details are discussed to provide a more thorough explanation of the embodiments of the present invention. However, it is obvious to those skilled in the art that the embodiments of the present invention can be implemented without these specific details. In other embodiments, well-known structures and devices are shown in the form of block diagrams rather than in detail to avoid making the embodiments of the present invention difficult to understand.

[0052] First of all, it should be noted that the corrosion fatigue behavior of aluminum alloys is affected by many characteristics, including material parameters (such as chemical composition, organization, heat treatment state, etc.), environmental parameters (such as corrosive media, concentration, temperature and humidity, pH, etc.) and mechanical parameters (such as stress ratio, frequency, waveform, etc.). Therefore, the corrosion fatigue failure of aluminum alloys is not just the superposition of simple corrosion and fatigue effects, but a multi-feature coupling effect under the synergistic effect of the interaction between corrosion factors and fatigue stress.

[0053] The present invention provides a training method, system, prediction method and device for a corrosion fatigue prediction model, which relates to the technical field of material corrosion fatigue life prediction, and can be specifically applied to solve the technical problems of the existing aluminum alloy corrosion fatigue prediction model, such as difficulty in describing influencing factors, lack of universality and low accuracy. The present invention can improve the prediction accuracy of aluminum alloy corrosion fatigue failure behavior and accurately predict its service life. The following is a detailed description through specific embodiments.

[0054] See also Figure 1 The training method of the corrosion fatigue prediction model provided by an embodiment of the present invention may include the following steps:

[0055] Step S100, obtaining a plurality of initial sample data of aluminum alloy corrosion fatigue and establishing an initial data set, wherein the initial sample data includes a plurality of characteristic values ​​and corresponding true values ​​of corrosion fatigue crack growth rates;

[0056] Step S200, inputting part of the characteristic values ​​of each initial sample data into a preset parameterized physical model to generate a corresponding prediction value of the corrosion fatigue crack growth rate;

[0057] Step S300, constructing a sample data set based on all characteristic values ​​of each initial sample data in the initial data set and the ratio between the corresponding true value and the predicted value of the corrosion fatigue crack growth rate;

[0058] Step S400: Use the sample data set to train the initial prediction model to generate an aluminum alloy corrosion fatigue prediction model.

[0059] In one embodiment of the present invention, when executing step S100, a plurality of initial sample data of aluminum alloy corrosion fatigue are obtained to establish an initial data set, wherein the initial sample data includes a plurality of characteristic values ​​and corresponding true values ​​of corrosion fatigue crack growth rates. Specifically, the following steps may be included:

[0060] Step S110, obtaining a plurality of initial sample data of aluminum alloy corrosion fatigue test;

[0061] Step S120: randomly generate multiple sets of feature values, input the multiple sets of feature values ​​into a preset sample expansion model, generate corresponding true values, and form multiple expanded initial sample data based on the multiple sets of feature values ​​and the corresponding true values;

[0062] Step S130: establishing an initial data set based on the expanded initial sample data and the acquired multiple initial sample data.

[0063] In one embodiment of the present invention, when executing step S110 to step S130, specifically, first, the initial sample data obtained from the relevant test is extracted from the laboratory internal test data, the public database or the published research paper. Then, the existing crack growth rate prediction model is used to generate additional initial sample data. In this embodiment, multiple different types of crack growth rate prediction models can be selected as preset sample extension models, such as the Paris equation, the Walker model, the Forman model and the NASGRO equation. In this embodiment, first, multiple groups of eigenvalues ​​are randomly generated within the effective range of the model, and each group of eigenvalues ​​is input into the preset sample extension model to generate corresponding true values. Then, multiple extended initial sample data are formed based on the multiple groups of eigenvalues ​​and the corresponding true values. According to the above steps, the extended initial sample data can be generated in batches, thereby obtaining more training samples. The above prediction model needs to meet the requirements of modeling, mathematization and high accuracy. Generating additional training samples through the existing prediction model can increase the diversity of data, so that more general patterns can be learned during subsequent model training, reducing dependence on specific samples, thereby reducing the risk of overfitting. Based on the sample data obtained from the experiment and the sample data obtained from the prediction, an initial data set is constructed.

[0064] In this embodiment, the initial data set may include multiple initial sample data, each of which includes multiple characteristic variables and the true values ​​of the aluminum alloy corrosion fatigue crack growth rate corresponding to these characteristic variables. The initial data set can be expressed as In each initial sample data middle, x i represents the i-th initial sample data containing z feature variables, It is represented as the target value corresponding to the i-th initial sample data, that is, the true value of the corrosion fatigue crack growth rate of aluminum alloy.

[0065] In one embodiment of the present invention, when executing step S200, all initial sample data are input into a preset parameterized physical model to obtain a predicted value of the corrosion fatigue crack growth rate corresponding to each initial sample data, and the parameterized physical model is associated with some characteristic variables of the initial sample data. Specifically, all initial sample data in the initial data set are sequentially input into the preset parameterized physical model, and the data input each time is the value of some modeling features associated with the parameterized physical model, so as to obtain a predicted value of the aluminum alloy corrosion fatigue crack growth rate corresponding to each initial sample data. In this embodiment, the preset parameterized physical model is obtained by the following steps:

[0066] Step S210, obtaining the yield strength in the material parameters of the aluminum alloy, the temperature in the environmental parameters, the stress intensity factor range in the mechanical parameters, the loading frequency in the mechanical parameters, and the stress ratio in the mechanical parameters as modeling features;

[0067] Step S220: construct a parameterized physical model based on multiple modeling features.

[0068] In one embodiment of the present invention, when executing step S210 to step S220, specifically, first, for the prediction problem of aluminum alloy corrosion fatigue crack growth rate, combined with the fatigue crack growth rate prediction equation of fracture mechanics (Paris equation, Walker equation, Forman equation, etc.), EPRI material reliability project stress corrosion crack growth rate model (MRP-386), aluminum alloy corrosion fatigue related influencing factors industry research results, etc., multiple physical information layer characteristic variables are selected from the three aspects of material parameters, environmental parameters and mechanical parameters as modeling features. For example, the yield strength (YS) is selected as the modeling feature for the material parameter, the temperature (T) is selected as the modeling feature for the environmental parameter, and the stress intensity factor range (ΔK), loading frequency (f) and stress ratio (R) are selected as the modeling features for the mechanical parameter.

[0069] Based on the modeling features of multiple selected physical information layers, a parameterized physical model for predicting the corrosion fatigue crack growth rate of aluminum alloy is constructed. In this embodiment, the following parameterized model can be obtained by extracting prior information of parameterized models such as the Paris equation, the Walker equation, and the Forman equation:

[0070] The parameterized model of material parameters is:

[0071] f YS =f (屈服强度) =YS γ ;

[0072] Among them, YS is the yield strength, and γ is the material constant, which indicates the influence of the yield strength YS on the crack growth rate.

[0073] The parameterized model of environmental parameters is:

[0074] f T =f (温度) =exp(-1 / RT);

[0075] Where R is the gas constant, which represents the effect of temperature on the crack growth rate, and T is the temperature.

[0076] The parameterized model of mechanical parameters is:

[0077] f N =f(ΔK，频率，应力比) =ΔK m (1 / f(1-R));

[0078] Among them, ΔK is the stress intensity factor range, m is the sensitivity index of the stress intensity factor range, f is the loading frequency, and R represents the stress ratio.

[0079] Then, based on the parameterized models of the three aspects of material parameters, environmental parameters and mechanical parameters, a parameterized physical model for predicting the corrosion fatigue crack growth rate of aluminum alloy is established. KN (x i ) satisfies the following formula:

[0080] f KN (x i ) = f YS ×f T ×f N

[0081] Among them, f YS is a parametric model of material parameters, and its modeling characteristics are yield strength YS, f T is a parameterized model of environmental parameters, whose modeling characteristics are temperature T, f N It is a parametric model of mechanical parameters, and its modeling characteristics are stress intensity factor range ΔK, loading frequency f and stress ratio R.

[0082] Next, each initial sample data in the initial data set is input into the parameterized physical model. The parameterized physical model can predict the corrosion fatigue performance of each initial sample data under specific conditions based on known physical laws and material properties, and output the corresponding crack growth rate prediction value. In this embodiment, the above process can be described by the following formula:

[0083]

[0084] From the above, we can see that the parameterized physical model only associates some modeling features of each initial sample data, that is, the model may not be able to fully capture all influencing factors, so there may be a certain deviation between the predicted value it generates and the true value.

[0085] In one embodiment of the present invention, when executing step S300, a sample data set is constructed based on all characteristic variables of each initial sample data in the initial data set and the ratio between the actual value and the predicted value of the corresponding corrosion fatigue crack growth rate. Specifically, after obtaining the predicted value of the corrosion fatigue crack growth rate of each initial sample data, the actual value can be calculated. Predicted values ​​from parameterized physical model output The ratio ci This ratio C i It reflects the difference between the predicted result of the parameterized physical model and the actual situation. i Satisfies the following formula:

[0086]

[0087] Then, based on all the feature variables x in each initial sample data i The corresponding ratio C i , construct a sample data set H i , the sample data set can be recorded as H i ={(x1,C1),(x2,C2),……(x i ,C i )}.

[0088] In one embodiment of the present invention, when executing step S400, the sample data set is used to train the initial prediction model, and the optimal parameters of the initial prediction model are calculated to generate an aluminum alloy corrosion fatigue prediction model. Specifically, the following steps may be included:

[0089] Step S410: using the XGBoost model as the initial prediction model, and setting the hyperparameters of the initial prediction model;

[0090] Step S420: input the sample data set into the initial prediction model for training, and perform hyperparameter optimization to generate an aluminum alloy corrosion fatigue prediction model.

[0091] In one embodiment of the present invention, when step S410 is executed, the XGBoost model is used as the initial prediction model, and the hyperparameters of the initial prediction model are set. Specifically, the present invention uses the XGBoost model as the initial prediction model. XGBoost (eXtreme Gradient Boosting) is a machine learning algorithm based on gradient boosting decision tree (GBDT), which can use CPU multi-threading for efficient parallel computing, has nonlinear segmentation capabilities and good overfitting control mechanisms. Compared with traditional machine learning algorithms, the XGBoost model has significant advantages in operations for large samples and high-dimensional data sets, and can effectively improve the model operation efficiency. The aluminum alloy corrosion fatigue problem has high-dimensional and complex characteristics and has good applicability. Therefore, this study uses the XGBoost model as a data-driven model. In this embodiment, the hyperparameters of the initial prediction model include the maximum depth of the tree, the learning rate, the subsample ratio, etc.

[0092] In one embodiment of the present invention, when executing step S420, the sample data set is input into the initial prediction model for training, and hyperparameter optimization is performed to generate an aluminum alloy corrosion fatigue prediction model. Specifically, the following steps may be included:

[0093] Step S421, dividing the sample data set into a training set and a test set, and using the training set to train the initial prediction model;

[0094] Step S422: Use the test set to evaluate the trained initial prediction model to generate a correlation coefficient of the initial prediction model;

[0095] Step S423: when the correlation coefficient is lower than the preset threshold, adjust the hyperparameters of the initial prediction model, retrain the adjusted initial prediction model using the training set, and repeatedly evaluate using the test set until the correlation coefficient reaches the preset threshold;

[0096] Step S424: when the correlation coefficient reaches a preset threshold, the trained initial prediction model is set as an aluminum alloy corrosion fatigue prediction model.

[0097] In one embodiment of the present invention, when executing steps S421 to S424, specifically, first, the initial data set is divided into a training set and a test set. The training set is used to train the model so that the model can learn the patterns and rules in the data. The test set is used to evaluate the generalization ability of the model, that is, the performance of the model on unseen data. First, the initial prediction model is trained using the training set; then, the trained initial prediction model is evaluated using the test set data. Evaluation indicators such as root mean square error (RMSE), mean absolute error (MAE), etc. are calculated. In this embodiment, the correlation coefficient (Coefficient of Determination, R 2 ) as the evaluation index. Correlation coefficient R 2 Satisfies the following formula:

[0098]

[0099] Among them, y i is the true value of the i-th sample, is the predicted value of the i-th sample, is the average value of all true values ​​of samples, and n is the number of all samples. In this embodiment, if the correlation coefficient R 2 The larger it is and the closer it is to 1, the better the prediction effect of the trained model. If the model accuracy meets the requirements, it means that the model has a good fitting effect; if the model accuracy does not meet the requirements, the model structure and hyperparameters are readjusted and retrained until the model meets the set accuracy requirements.

[0100] In this embodiment, model training is performed according to step S410 to step S420, and an aluminum alloy corrosion fatigue prediction model that meets the accuracy requirements is obtained, thereby establishing a sample size x i With ratio C i The parameterized physical model with specific domain knowledge is further combined with the XGBoost machine learning model to establish a mechanism analysis and data-driven coupling model as follows:

[0101]

[0102] It can be seen from the above formula that when predicting aluminum alloy corrosion fatigue, multiple characteristic values ​​of the aluminum alloy material to be predicted can be obtained first. Then, the multiple characteristic values ​​are input into the trained aluminum alloy corrosion fatigue prediction model for prediction to generate the corresponding prediction ratio C i Finally, some of the multiple eigenvalues ​​are input into the preset parameterized physical model to generate the corresponding prediction value f KN (x i ), and based on the predicted ratio C i and the predicted value f KN (x i ), calculate and generate the corrosion fatigue crack growth rate of aluminum alloy By combining prior knowledge with data-driven modeling results, the present invention can better learn the physical mechanism of crack propagation and improve the prediction model's accuracy, generalization performance and interpretability.

[0103] See also Figure 2 In one embodiment of the present invention, 280 aluminum alloy corrosion fatigue test data were collected through literature to establish an initial data set. After data enhancement by multiple existing prediction models, the number of samples in the initial data set was expanded to 640, so the corresponding sample data set was also 640. The training set can be set to 500 training samples, and the test set can be set to 140 test samples. The training set is used to train the aluminum alloy corrosion fatigue crack growth rate prediction model that integrates the established domain knowledge and data-driven, and the test set verifies the model effect, and finally the R of the model is obtained. 2 The fitting effect is good.

[0104] See also Figure 3 Random new data are used to predict the extension of the aluminum alloy corrosion fatigue prediction model. The prediction trend of the coupled model of the present invention is more in line with the research rules of the industry than the data-driven model. The addition of prior knowledge can help the model better learn the physical mechanism of crack propagation and improve the generalization performance and interpretability of the model.

[0105] See also Figure 4 The present invention also provides a training system for a corrosion fatigue prediction model, which corresponds to the training method in the above embodiment. The training system may include a data acquisition module 11, an initial prediction module 12, a sample construction module 13 and a model training module 14. The functional modules are described in detail as follows:

[0106] The data acquisition module 11 can be used to acquire multiple initial sample data of aluminum alloy corrosion fatigue and establish an initial data set, wherein the initial sample data includes multiple eigenvalues ​​and corresponding true values ​​of corrosion fatigue crack growth rate. Further, the data acquisition module 11 can be specifically used to randomly generate multiple groups of eigenvalues, and input each group of eigenvalues ​​into a preset sample expansion model to generate corresponding true values, and form multiple extended initial sample data based on the multiple groups of eigenvalues ​​and the corresponding true values; establish an initial data set based on the extended initial sample data and the multiple initial sample data of the aluminum alloy corrosion fatigue test acquired.

[0107] The initial prediction module 12 can be used to input part of the characteristic values ​​of each initial sample data into a preset parameterized physical model to generate a corresponding prediction value of the corrosion fatigue crack growth rate. Further, the initial prediction module 12 can be specifically used to input all the initial sample data in the initial data set into the preset parameterized physical model in sequence, and the data input each time is the value of part of the modeling characteristics associated with the parameterized physical model, so as to obtain the prediction value of the aluminum alloy corrosion fatigue crack growth rate corresponding to each initial sample data.

[0108] The sample construction module 13 can be used to construct a sample data set based on all the characteristic values ​​of each initial sample data in the initial data set and the ratio between the actual value and the predicted value of the corresponding corrosion fatigue crack growth rate. Further, the sample construction module 13 can be specifically used to calculate the actual value after obtaining the predicted value of the corrosion fatigue crack growth rate of each initial sample data. Predicted values ​​from parameterized physical model output The ratio C i This ratio C i It reflects the difference between the predicted results of the parameterized physical model and the actual situation. Then, based on all the eigenvalues ​​of each initial sample data in the initial data set and the ratio between the actual value and the predicted value of the corresponding corrosion fatigue crack growth rate, a sample data set is constructed.

[0109] The model training module 14 can be used to train the initial prediction model using the sample data set to generate an aluminum alloy corrosion fatigue prediction model. Further, the model training module 14 can be specifically used to use the XGBoost model as the initial prediction model and set the hyperparameters of the initial prediction model; input the sample data set into the initial prediction model for training, and perform hyperparameter optimization to generate an aluminum alloy corrosion fatigue prediction model.

[0110] The specific definition of the training system of the corrosion fatigue prediction model can be found in the definition of the training method above, which will not be repeated here. Each module in the above training system can be implemented in whole or in part by software, hardware and a combination thereof. The above modules can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory of the computer device in the form of software, so that the processor can call and execute the operations corresponding to the above modules.

[0111] The present invention also provides a method for predicting aluminum alloy corrosion fatigue, comprising: obtaining multiple characteristic values ​​of the aluminum alloy material to be predicted; inputting the multiple characteristic values ​​into a prediction model for prediction, generating corresponding prediction ratios; inputting some characteristic values ​​of the multiple characteristic values ​​into a preset parameterized physical model, generating corresponding prediction values, and calculating and generating the aluminum alloy corrosion fatigue crack growth rate based on the prediction ratio and the prediction value. In this embodiment, the prediction model is the aluminum alloy corrosion fatigue prediction model trained by any of the above embodiments, and the parameterized physical model is also the parameterized physical model described in the above embodiments.

[0112] In this embodiment, the aluminum alloy corrosion fatigue crack growth rate Satisfies the following formula:

[0113]

[0114] Among them, x i represents multiple eigenvalues, f KN (x i ) represents the predicted value of the parameterized physical model output, C i Represents the predicted ratio.

[0115] An embodiment of the present invention also provides an electronic device, comprising: one or more processors; a storage device for storing one or more programs, wherein when the one or more programs are executed by one or more processors, the electronic device implements the training method of the corrosion fatigue prediction model or the prediction method of aluminum alloy corrosion fatigue provided in the above-mentioned embodiments.

[0116] See also Figure 5The electronic device 2 may include a memory 22, a processor 23 and a bus, and may also include a computer program stored in the memory 22 and executable on the processor 23, such as a training program for a corrosion fatigue prediction model or a prediction program for aluminum alloy corrosion fatigue.

[0117] Among them, the memory 22 includes at least one type of readable storage medium, and the readable storage medium includes flash memory, mobile hard disk, multimedia card, card-type memory (for example: SD or DX memory, etc.), magnetic memory, disk, optical disk, etc. In some embodiments, the memory 22 can be an internal storage unit of the electronic device 2, such as a mobile hard disk of the electronic device 2. In other embodiments, the memory 22 can also be an external storage device of the electronic device 2, such as a plug-in mobile hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), etc. equipped on the electronic device 2. Further, the memory 22 can also include both an internal storage unit of the electronic device 2 and an external storage device. The memory 22 can not only be used to store application software and various types of data installed in the electronic device 2, such as the code for training the corrosion fatigue prediction model or the code for predicting the corrosion fatigue of aluminum alloys, but can also be used to temporarily store data that has been output or is to be output.

[0118] In some embodiments, the processor 23 may be composed of an integrated circuit, for example, a single packaged integrated circuit, or a plurality of packaged integrated circuits with the same or different functions, including one or more central processing units (CPUs), microprocessors, digital processing chips, graphics processors, and combinations of various control chips. The processor 23 is the control core (Control Unit) of the electronic device 2, and uses various interfaces and lines to connect various components of the entire electronic device 2, and executes or executes programs or modules stored in the memory 22 (such as a training program for a corrosion fatigue prediction model, etc.), and calls data stored in the memory 22 to execute various functions of the electronic device 2 and process data.

[0119] The processor 23 executes the operating system and various installed applications of the electronic device 2. The processor 23 executes the applications to implement the steps in the training method of the corrosion fatigue prediction model or the steps in the prediction method of aluminum alloy corrosion fatigue.

[0120] Exemplarily, the computer program may be divided into one or more modules, which are stored in the memory 22 and executed by the processor 23 to complete the present application. The one or more modules may be a series of computer program instruction segments capable of completing specific functions, which are used to describe the execution process of the computer program in the electronic device 2. For example, the computer program may be divided into a data acquisition module 11, an initial prediction module 12, a sample construction module 13, and a model training module 14.

[0121] The above-mentioned integrated unit implemented in the form of a software function module can be stored in a computer-readable storage medium, and the computer-readable storage medium can be non-volatile or volatile. The above-mentioned software function module is stored in a storage medium, including a number of instructions for enabling a computer device (which can be a personal computer, a computer device, or a network device, etc.) or a processor to perform part of the functions of the training method of the corrosion fatigue prediction model or the prediction method of aluminum alloy corrosion fatigue described in various embodiments of the present application.

[0122] In summary, the present invention discloses a training method, system, prediction method and device for a corrosion fatigue prediction model, which is based on a method for predicting the corrosion fatigue crack growth rate of aluminum alloys that integrates prior knowledge and data-driven, and can use the modeled and mathematized basic knowledge, laws or experience of corrosion fatigue to guide specific learning algorithms to improve learning performance. After data verification, the addition of prior knowledge can help the model better learn the physical mechanism of crack propagation, improve the model's accuracy, generalization performance and interpretability, and is beneficial to the safety of aluminum alloy corrosion fatigue behavior prediction engineering applications. Therefore, the present invention effectively overcomes the various shortcomings of the prior art and has a high industrial utilization value.

[0123] The above embodiments are merely illustrative of the principles and effects of the present invention, and are not intended to limit the present invention. Anyone familiar with the art may modify or alter the above embodiments without departing from the spirit and scope of the present invention. Therefore, all equivalent modifications or alterations made by a person of ordinary skill in the art without departing from the spirit and technical concept disclosed by the present invention shall still be covered by the claims of the present invention.

Claims

1. A training method for a corrosion fatigue prediction model, characterized in that: include: Acquire multiple initial sample data of aluminum alloy corrosion fatigue and establish an initial data set, wherein the initial sample data includes multiple characteristic values ​​and corresponding true values ​​of corrosion fatigue crack growth rates; Inputting partial characteristic values ​​of each initial sample data into a preset parameterized physical model to generate a corresponding prediction value of the corrosion fatigue crack growth rate; Constructing a sample data set based on all characteristic values ​​of each initial sample data in the initial data set and the ratio between the corresponding true value and the predicted value of the corrosion fatigue crack growth rate; The sample data set is used to train the initial prediction model to generate an aluminum alloy corrosion fatigue prediction model.

2. The training method of the corrosion fatigue prediction model according to claim 1 is characterized in that: The step of obtaining a plurality of initial sample data of aluminum alloy corrosion fatigue and establishing an initial data set, wherein the initial sample data includes a plurality of characteristic values ​​and corresponding true values ​​of corrosion fatigue crack growth rates, comprises: Multiple initial sample data obtained for aluminum alloy corrosion fatigue testing; Randomly generate multiple sets of eigenvalues, input the multiple sets of eigenvalues ​​into a preset sample expansion model, generate corresponding true values, and form multiple expanded initial sample data based on the multiple sets of eigenvalues ​​and the corresponding true values; An initial data set is established based on the expanded initial sample data and the acquired multiple initial sample data.

3. The training method of the corrosion fatigue prediction model according to claim 1 is characterized in that: The parameterized physical model is generated according to the following steps: The yield strength in the material parameters of the aluminum alloy, the temperature in the environmental parameters, the range of the stress intensity factor in the mechanical parameters, the loading frequency in the mechanical parameters, and the stress ratio in the mechanical parameters are obtained as modeling features; Based on a plurality of the modeling features, a parameterized physical model is constructed.

4. The training method of the corrosion fatigue prediction model according to claim 3 is characterized in that: The parameterized physical model f KN (x i ) satisfies the following formula: f KN (x i )=f YS ×f T ×f N Among them, x i represents the initial sample containing multiple eigenvalues, f Ys represents the parameterized model of yield strength in terms of material parameters, f T represents the parameterized model about the temperature in the environmental parameters, f N Represents a parameterized model with respect to the range of stress intensity factors, loading frequency, and stress ratios in the mechanical parameters.

5. The training method of the corrosion fatigue prediction model according to claim 1, characterized in that: The step of using the sample data set to train the initial prediction model to generate an aluminum alloy corrosion fatigue prediction model comprises: Using the XGBoost model as an initial prediction model and setting hyperparameters of the initial prediction model; The sample data set is input into the initial prediction model for training, and hyperparameter optimization is performed to generate an aluminum alloy corrosion fatigue prediction model.

6. The training method of the corrosion fatigue prediction model according to claim 5, characterized in that: The step of inputting the sample data set into the initial prediction model for training and performing hyperparameter optimization to generate an aluminum alloy corrosion fatigue prediction model includes: Dividing the sample data set into a training set and a test set, and using the training set to train the initial prediction model; Using the test set to evaluate the trained initial prediction model, and generating a correlation coefficient of the initial prediction model; When the correlation coefficient is lower than a preset threshold, adjusting the hyperparameters of the initial prediction model, retraining the adjusted initial prediction model using the training set, and repeatedly evaluating using the test set until the correlation coefficient reaches the preset threshold; When the correlation coefficient reaches a preset threshold, the trained initial prediction model is set as the aluminum alloy corrosion fatigue prediction model.

7. A training system for a corrosion fatigue prediction model, characterized in that: include: A data acquisition module, used to acquire a plurality of initial sample data of aluminum alloy corrosion fatigue and establish an initial data set, wherein the initial sample data includes a plurality of characteristic values ​​and corresponding true values ​​of corrosion fatigue crack growth rates; An initial prediction module is used to input part of the characteristic values ​​of each initial sample data into a preset parameterized physical model to generate a corresponding prediction value of the corrosion fatigue crack growth rate; A sample construction module, used to construct a sample data set based on all characteristic values ​​of each initial sample data in the initial data set and the ratio between the actual value and the predicted value of the corresponding corrosion fatigue crack growth rate; The model training module is used to train the initial prediction model using the sample data set to generate an aluminum alloy corrosion fatigue prediction model.

8. A method for predicting corrosion fatigue of aluminum alloy, characterized in that: include: Acquire multiple characteristic values ​​of the aluminum alloy material to be predicted; Inputting a plurality of characteristic values ​​into a prediction model for prediction to generate corresponding prediction ratios, wherein the prediction model is the aluminum alloy corrosion fatigue prediction model according to any one of claims 1 to 6; Some of the multiple eigenvalues ​​are input into a preset parameterized physical model to generate corresponding prediction values, and based on the prediction ratio and the prediction value, the corrosion fatigue crack growth rate of the aluminum alloy is calculated and generated.

9. The method for predicting aluminum alloy corrosion fatigue according to claim 8, characterized in that: The aluminum alloy corrosion fatigue crack growth rate Satisfies the following formula: Among them, x i represents multiple eigenvalues, f KN (x i ) represents the predicted value of the parameterized physical model output, C i represents the predicted ratio.

10. An electronic device, characterized in that: include: one or more processors; A storage device for storing one or more programs, which, when executed by the one or more processors, enables the electronic device to implement the training method of the corrosion fatigue prediction model as described in any one of claims 1 to 6 or the prediction method of aluminum alloy corrosion fatigue as described in claims 8-9.

Images