A method for predicting fatigue crack growth life of TiAl alloy based on machine learning
By combining machine learning and linear elastic fracture mechanics, the problem of efficient and low-cost prediction of fatigue crack propagation life in TiAl alloys was solved, achieving high-precision crack propagation life prediction and reducing the risk of aero-engine blade fracture.
Patent Information
- Application Number
- CN202210886710.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2022-07-26
- Publication Date
- 2024-11-19
- Estimated Expiration
- 2042-07-26
AI Technical Summary
Existing technologies are insufficient for efficiently and cost-effectively predicting the fatigue crack propagation life of TiAl alloys, leading to a high risk of sudden fracture of aero-engine blades.
Using machine learning methods, a random forest model was constructed by acquiring data on effective stress intensity factor and fatigue crack propagation rate. Combined with linear elastic fracture mechanics theory, the fatigue crack propagation life of TiAl alloy was predicted.
This method enables low-cost, high-precision prediction of fatigue crack propagation life in TiAl alloys, reducing R&D and testing costs and improving prediction accuracy and efficiency.
Smart Images

Figure CN115169240B_ABST
Abstract
Description
Technical Field
[0001] The invention relates to a TiAl alloy fatigue crack extension life prediction method based on machine learning, and belongs to the field of material fatigue life prediction. Background Art
[0002] TiAl alloy has excellent creep resistance, oxidation resistance and high specific strength. It is the only lightweight high-temperature alloy material that can be used for a long time in an oxidizing environment above 600℃. It is currently widely used in hot end components of aerospace vehicles, such as aircraft engine blades. Fatigue is the main failure form of aircraft engine blades, and it is also the most dangerous failure form. When the fatigue life limit is reached, there is no obvious sign and it suddenly breaks and disintegrates. Therefore, predicting the fatigue crack growth life of TiAl alloy is of great significance to assess its service conditions.
[0003] In recent years, machine learning has become a frontier direction and hot area in materials research, and has significant advantages in prediction efficiency and cost control. Machine learning has excellent fitting ability for nonlinear data, so it can be mined and learned from the fatigue crack growth test of TiAl alloy to obtain the mapping relationship between effective stress intensity factor and fatigue crack growth rate. In addition, the machine learning model has good encapsulation, so based on the machine model and fatigue crack growth theory in linear elastic fracture mechanics, the crack growth rate can be converted into crack growth life, so as to conveniently, quickly and accurately predict the crack growth life curve of samples with different initial conditions.
[0004] Based on the above reasons, it is urgent to develop a TiAl alloy fatigue crack growth life prediction method based on machine learning to save R&D and testing costs. Summary of the invention
[0005] The purpose of the present invention is to provide a machine learning method for predicting the fatigue crack growth life of TiAl alloy, so as to achieve low-cost, high-precision and accurate prediction of the fatigue life of TiAl alloy.
[0006] In order to achieve the above object, the present invention provides a TiAl alloy fatigue crack growth life prediction method based on machine learning, which specifically comprises the following steps:
[0007] Step 1: Obtain the effective stress intensity factor ΔK of TiAl alloy eff and the corresponding fatigue crack growth rate da / dN as a set of data, and multiple sets of data constitute the original data set;
[0008] Step 2: numerically process the original data set in step 1;
[0009] Step 3: Randomly divide the data set processed in step 2 into a training set and a test set, build a machine learning model, set the parameter set, train the model with the training set, evaluate the trained model with the test set, and adjust the parameter set until the optimal model is obtained;
[0010] Step 4: Based on the fatigue crack growth theory in linear elastic fracture mechanics, the initial conditions of the TiAl alloy fatigue test to be predicted are input into the optimal model described in step 3 to obtain the predicted value of the fatigue crack growth life of the TiAl alloy.
[0011] Preferably, in step 2, the effective stress intensity factor ΔK in each set of data in the original data set is eff and the corresponding fatigue crack growth rate da / dN are processed logarithmically to obtain lg(ΔK eff ) and lg(da / dN).
[0012] Preferably, in step three, the processed data set is randomly divided into a training set and a test set, wherein 80% is the training set and 20% is the test set.
[0013] Preferably, in step three, the machine learning model constructed is a random forest model.
[0014] Preferably, in step three, the parameter set includes the number of decision trees n, the maximum depth of the decision tree d, and the maximum number of features of the decision tree f.
[0015] Preferably, in step 3, the regression coefficient R is used 2 And the root mean square error RMSE is used as the accuracy to evaluate the trained model.
[0016] Among them, the regression coefficient R 2 for:
[0017]
[0018] Where n is the number of data groups in the test set, y j is the crack growth rate in the test center, is the crack growth rate predicted by the model in the test set, is the average value of all crack growth rates in the test set;
[0019] The root mean square error is RMSE:
[0020]
[0021] Where n is the number of data groups in the test set, y j is the crack growth rate in the test center, Crack growth rates predicted by the model for the test set.
[0022] Preferably, in step three, a gradient descent algorithm is used to adjust the parameter set.
[0023] Preferably, in step 4, the specimen used in the fatigue test of the TiAl alloy to be predicted is a standard compact tensile specimen in "GB / T6398-2017 Fatigue Crack Propagation Method for Fatigue Test of Metallic Materials".
[0024] Preferably, in step 4, the initial conditions of the TiAl alloy fatigue test to be predicted are a force range ΔF, a sample thickness B, a sample width W, and an initial crack length a0.
[0025] Preferably, in step 4, the fatigue crack growth theory formula in linear elastic fracture mechanics is specifically:
[0026]
[0027] The shape factor is:
[0028]
[0029] Among them, α=a / W.
[0030] Compared with the prior art, the present invention has the following significant advantages:
[0031] The present invention can fully consider the nonlinear problem in the process of predicting the fatigue crack growth life of TiAl alloy, avoiding the calculation error caused by traditional linear fitting in the prediction of crack growth life; at the same time, it can conveniently, quickly and accurately predict the crack growth life curve of samples with different initial conditions, which can reduce the calculation cost to save research and development and test costs. BRIEF DESCRIPTION OF THE DRAWINGS
[0032] Figure 1 This is a comparison chart of the crack growth rate predicted by the TiAl alloy fatigue crack growth life prediction method based on machine learning and the experimental value.
[0033] Figure 2 This is a fatigue crack growth life curve predicted by the TiAl alloy fatigue crack growth life prediction method based on machine learning of the present invention. DETAILED DESCRIPTION
[0034] The present invention is further described in detail below in conjunction with the embodiments and drawings.
[0035] Example
[0036] The data-model mutually driven TiAl alloy fatigue life prediction method described in the present invention specifically comprises the following steps:
[0037] Step 1: Obtain the effective stress intensity factor ΔK of TiAl alloy effand the corresponding fatigue crack growth rate da / dN as a set of data, a total of 49 sets of data constitute the original data set;
[0038] Step 2: The effective stress intensity factor ΔK in each data set of the original data set eff and the corresponding fatigue crack growth rate da / dN are processed logarithmically to obtain lg(ΔK eff ) and lg(da / dN);
[0039] Step 3: Randomly divide the data set processed in step 2 into training set and test set, 80% of which is the training set and 20% is the test set. Build a random forest model and set the parameter set, which includes the number of decision trees n, the maximum depth of the decision tree d, and the maximum number of features of the decision tree f. Use the training set to train the model and use the regression coefficient R for the test set. 2 And the root mean square error RMSE is used as the accuracy to evaluate the trained model.
[0040] Among them, the regression coefficient R 2 for:
[0041]
[0042] Where n is the number of data groups in the test set, y j is the crack growth rate in the test center, is the crack growth rate predicted by the model in the test set, is the average value of all crack growth rates in the test set;
[0043] The root mean square error is RMSE:
[0044]
[0045] Where n is the number of data groups in the test set, y j is the crack growth rate in the test center, is the crack growth rate predicted by the model in the test set,
[0046] The gradient descent algorithm is used to adjust the parameter set until the optimal model is obtained.
[0047] The adjusted test set is as follows: the number of decision trees n = 44, the maximum depth of the decision tree d = 9, the maximum number of features of the decision tree f = 1, and the regression coefficient R of the optimal model 2 =0.935 The root mean square error is RMSE = 0.296. The crack growth rate predicted by the model is compared with the experimental value. Figure 1 As shown;
[0048] Step 4: Based on the fatigue crack growth theory in linear elastic fracture mechanics, the sample is the standard compact tensile specimen in "GB / T6398-2017 Fatigue Crack Growth Method for Fatigue Test of Metallic Materials", and the initial conditions of the fatigue test of the TiAl alloy to be predicted: the force range ΔF is 790.57N, 869.63N, 948.68N, the sample thickness B = 0.05m, the sample width W = 0.25m, and the initial crack length a0 = 0.15m are input into the optimal model described in step 3 to obtain the predicted value of the fatigue crack growth life of the TiAl alloy. The calculation process is as follows:
[0049] According to the fatigue crack growth theory formula in linear elastic fracture mechanics:
[0050]
[0051] The shape factor is:
[0052]
[0053] Where, α=a / W,
[0054] Since the initial conditions ΔF, B, and W are given, substituting them into the formula yields the effective stress intensity factor ΔK eff The function of crack length a is:
[0055] ΔK eff =h(a),
[0056] The optimal model obtained in step 3 is essentially the fatigue crack growth rate da / dN to the effective stress intensity factor ΔK eff The mapping is:
[0057]
[0058] therefore:
[0059]
[0060] It can be roughly considered that:
[0061]
[0062] Therefore, we can get it by iteration:
[0063]
[0064] Since a0 is given, we can set the iteration step size ΔN = N i+1 -N i =1, we can get a1, a2, a3, ..., a n, thus obtaining the aN curve, that is, the fatigue crack growth life curve, such as Figure 2 shown.
Claims
1. A method for predicting fatigue crack growth life of TiAl alloy based on machine learning, characterized in that: The specific steps include: Step 1: Obtain the effective stress intensity factor ΔK of TiAl alloy eff and the corresponding fatigue crack growth rate da / dN as a set of data, and multiple sets of data constitute the original data set; Step 2: numerically process the original data set in step 1; Step 3: Randomly divide the data set processed in step 2 into a training set and a test set, build a machine learning model, set the parameter set, train the model with the training set, evaluate the trained model with the test set, and adjust the parameter set until the optimal model is obtained; Step 4: Based on the fatigue crack growth theory in linear elastic fracture mechanics, the initial conditions of the fatigue test of the TiAl alloy to be predicted are input into the optimal model described in step 3 to obtain the predicted value of the fatigue crack growth life of the TiAl alloy; Among them, in step three, the machine learning model constructed is a random forest model.
2. The method according to claim 1, characterized in that In step 2, the effective stress intensity factor ΔK in each data set of the original data set is eff and the corresponding fatigue crack growth rate da / dN are processed logarithmically to obtain lg(ΔK eff ) and lg(da / dN).
3. The method according to claim 1, characterized in that In step three, the processed data set is randomly divided into a training set and a test set, with 80% being the training set and 20% being the test set.
4. The method according to claim 1, characterized in that In step 3, the parameter set includes the number of decision trees n, the maximum depth of the decision tree d, and the maximum number of features of the decision tree f.
5. The method according to claim 1, characterized in that In step 3, the regression coefficient R 2 And the root mean square error RMSE is used as the accuracy to evaluate the trained model. Among them, the regression coefficient R 2 for: Where n is the number of data groups in the test set, y j is the crack growth rate in the test center, is the crack growth rate predicted by the model in the test set, is the average value of all crack growth rates in the test set; The root mean square error is RMSE: Where n is the number of data groups in the test set, y j is the crack growth rate in the test center, Crack growth rates predicted by the model for the test set.
6. The method according to claim 1, characterized in that In step three, the gradient descent algorithm is used to adjust the parameter set.
7. The method according to claim 1, characterized in that In step 4, the initial conditions of the TiAl alloy fatigue test to be predicted are the force range ΔF, the sample thickness B, the sample width W, and the initial crack length a0.
8. The method according to claim 7, characterized in that In step 4, the fatigue crack growth theory formula in linear elastic fracture mechanics is as follows: The shape factor is: Among them, α=a / W.
