CN-122024947-A - Fatigue crack growth prediction method and system based on incremental information machine learning modeling

CN122024947ACN 122024947 ACN122024947 ACN 122024947ACN-122024947-A

Abstract

The invention provides a fatigue crack growth prediction method and a system based on incremental information machine learning modeling, and relates to the technical field of fatigue crack growth prediction. Firstly, collecting early data samples of a fatigue crack expansion test of a sample as an original data set, and performing iterative expansion of the data samples through machine learning modeling interpolation prediction to obtain an expanded data set. Based on the extended data set, an incremental information sample data set is constructed. And carrying out secondary machine learning modeling to obtain a fatigue crack growth rate prediction model. Finally, a fatigue crack growth rate prediction model is utilized, and a crack length value and a confidence interval thereof in the later stage of the test of the fatigue crack growth of the sample are obtained through multi-path prediction weighted average, so that a complete fatigue crack growth curve of the sample is finally obtained. According to the method, the high-efficiency and accurate prediction of the fatigue crack expansion curve is realized by interpolation prediction data expansion and machine learning modeling based on the incremental information and by fusion of the incremental information.

Inventors

TIAN YUWAN
WANG YILIANG
WEN CHENG
GAO KEWEI
LU HAIPENG
YAN CHANGJIAN
DIAO YUPENG

Assignees

广东海洋大学
广东腐蚀科学与技术创新研究院

Dates

Publication Date: 20260512
Application Date: 20260105

Claims (10)

1. The fatigue crack growth prediction method based on incremental information machine learning modeling is characterized by comprising the following steps of: S1, collecting early data samples of a fatigue crack propagation test of a sample, and taking the early data samples as an original data set; S2, based on an original data set, performing iterative expansion of a data sample through machine learning modeling interpolation prediction to obtain an expanded data set; S3, constructing an increment information sample data set containing cyclic cycle increment and crack length increment based on the expansion data set; S4, performing secondary machine learning modeling based on the increment information sample data set to obtain a fatigue crack growth rate prediction model; S5, obtaining a crack length value and a confidence interval thereof at the later stage of the test of the fatigue crack growth of the sample by utilizing a fatigue crack growth rate prediction model and through multipath prediction weighted average, and obtaining a complete fatigue crack growth curve of the sample.
2. The method for predicting fatigue crack growth based on incremental information machine learning modeling of claim 1, wherein the step of collecting early data samples of the specimen fatigue crack growth test as the raw data set comprises the steps of: And acquiring early data of a fatigue crack propagation curve of the sample through a fatigue crack propagation test to obtain an original data set, wherein the expression of the original data set is as follows: D n ={(a 1 ,N 1 ), (a 2 ,N 2 ), ..., (a n ,N n )} where N represents the number of data samples in the original dataset, N n represents the cycle of the fatigue test, a n represents the corresponding crack length at N n cycles; The raw data set is collected until a n reaches half of the fatigue crack growth length a before the sample breaks, and a raw data set D n is obtained.
3. The fatigue crack growth prediction method based on incremental information machine learning modeling of claim 1, wherein the process of step S2 is: Firstly, constructing a neural network model, wherein the neural network model takes the cycle number N as input and the crack length a as output; Training the neural network model based on an original data set of a sample fatigue crack propagation test to obtain a trained neural network model; Dividing the cyclic cycle range [ N 1 ,N n ] of the original data set into P equidistant intervals to obtain P n cyclic cycle nodes, wherein the initial p=2; Predicting the crack length corresponding to each cycle node by using the trained neural network model to obtain an extended data set D 2n , and simultaneously obtaining a crack length prediction error E of the neural network model to the original data set; Step five, order Retraining by using the extended data set D 2n to update the neural network model, repeating the steps three to four, and iteratively updating the extended data set D 3n ,D 4n based on interpolation prediction until the fluctuation of the crack length prediction error E of the neural network model on the original data set is smaller than a preset error value, thereby obtaining the extended data set D pn , wherein the expression is as follows: D pn ={(a 1 ,N 1 ), (a 2 ,N 2 ), ... ,(a i ,N i ),..., (a j ,N j ),..., (a pn ,N pn )} Where D pn represents the extended dataset.
4. The fatigue crack growth prediction method based on incremental information machine learning modeling according to claim 1, wherein the step S3 specifically comprises: Based on the extended data set D pn , calculating the cycle increment delta N ij =N j -N i and the crack length increment delta a ij =a j -a i among all sample points in the extended data set, wherein j > i, obtaining the crack expansion rate delta a ij /ΔN ij among all sample points, constructing an increment information sample data set S, and the expression is as follows S={ (N 1 , ΔN 12 , Δa 12 /ΔN 12 ), (N 1 , ΔN 13 , Δa 13 /ΔN 13 ), ... ,(N i , ΔN ij , Δa ij /ΔN ij ),..., (N (pn-1) , ΔN (pn-1)pn , Δa (pn-1)pn /ΔN (pn-1)pn ) } The incremental information sample dataset S includes input features including cycle N i and cycle interval DeltaN ij , and output targets that are crack growth rates Deltaa ij /ΔN ij .
5. The fatigue crack growth prediction method based on incremental information machine learning modeling of claim 1, wherein the step S4 is to construct a single hidden layer feedforward neural network model, train the single hidden layer feedforward neural network model based on the incremental information sample data set S, and use a mean square error as a loss function until the loss function converges in the training process to obtain a fatigue crack growth rate prediction model.
6. The method for predicting fatigue crack growth based on incremental information machine learning modeling of claim 5, wherein step S5 comprises obtaining an average crack growth rate using a fatigue crack growth rate prediction model, comprising: Step one, taking the cycle interval of adjacent sample points in the D pn expansion data set as the step length delta N, constructing a corresponding cycle set of the later stage of the fatigue crack expansion test of the sample [ N n＋1 , N n＋2 ,..., N n＋i ,..., N 2pn ], Wherein N n＋2 -N n＋1 = Δn; Calculating cycle intervals of each cycle in the cycle set and each sample point in the D n dataset respectively, and calculating a general formula of delta N 1(n＋i) , ΔN 2(n＋i) , ... , ΔN i(n＋i) , ..., ΔN j(n＋i) , ..., ΔN n(n＋i) for N n＋i, ; Step three, constructing and obtaining a data set for predicting the crack length of the corresponding cycle in the later stage of the fatigue crack growth test of the sample, and calculating the general formula of N n＋i, as follows ：(N 1 , ΔN 1(n＋i) ), (N 2 , ΔN 2(n＋i) ), ... , (N i , ΔN i(n＋i) ), ... , (N j , ΔN j(n＋i) ), ... , (N n , ΔN n(n＋i) ); Step four, inputting the data set in the step three into a trained fatigue crack growth rate prediction model to obtain the fatigue crack growth rate of the sample under the corresponding cycle of the later fatigue crack growth test, Calculate the general formula for N n＋i, ：Δa 1(n＋i) /ΔN 1(n＋i) , Δa 2(n＋i) /ΔN 2(n＋i) , ... , Δa i(n＋i) /ΔN i(n＋i) , ... , Δa j(n＋i) /ΔN j(n＋i) , ... , Δa n(n＋i) /ΔN n(n＋i) .
7. The fatigue crack growth prediction method based on incremental information machine learning modeling according to claim 1, wherein the fatigue crack growth prediction method is characterized in that a crack length value and a confidence interval thereof in the later stage of a sample fatigue crack growth test are obtained through multipath prediction weighted average, and a complete fatigue crack growth curve of the sample is obtained through the following specific processes: Based on n results of fatigue crack growth rate prediction under the corresponding cycle of the later test period of the fatigue crack growth of the sample, n fatigue crack lengths under the corresponding cycle of the later test period of the fatigue crack growth of the sample are calculated: The general formula was calculated for N n＋i, : integrating n fatigue crack lengths, and obtaining crack length values and confidence intervals thereof in the later stage of the fatigue crack expansion test of the sample through multipath prediction weighted average: Aiming at N fatigue crack lengths of cycle N n＋i { a 1(n＋i) ,a 2(n＋i) ,...,a n(n＋i) }, calculating a weighted average of predicted crack lengths based on a predicted crack length set of target predicted points to obtain a fatigue crack expansion curve with complete samples, wherein the weighted average is as follows: Based on the delta Δa i,n＋i for each predicted crack length, a path length weight is calculated, expressed as: Wherein ω i represents the weight of the ith path, Δa k,n+i represents the crack length increment of the kth path, and n represents the number of fixed reference points; calculating the final predicted crack length, wherein the expression is: wherein a n+i,final represents the final predicted crack length corresponding to the target predicted point N n+i ; Calculating a confidence interval, wherein the expression is as follows: In the formula, sd (a n+i ) represents the standard deviation of the predicted crack length corresponding to the target predicted point N n+i , that is, the confidence interval.
8. The method for predicting fatigue crack growth based on incremental information machine learning modeling of claim 7, further comprising sequentially connecting fatigue cracks of the sample under corresponding cycles at early and late stages of the test for fatigue crack growth to obtain a complete fatigue crack growth curve of the sample.
9. An electronic device comprising a memory, a processor and a computer program stored on the memory and running on the processor, the incremental information machine learning modeling based fatigue crack growth prediction program implementing the steps of the incremental information machine learning modeling based fatigue crack growth prediction method of any of claims 1-8 when executed by the processor.
10. A fatigue crack growth prediction system based on incremental information machine learning modeling for implementing the incremental information machine learning modeling-based fatigue crack growth prediction method of any one of claims 1-8, comprising: the data acquisition module is used for acquiring early data samples of the fatigue crack propagation test of the sample and taking the early data samples as an original data set; the data expansion module is used for carrying out iterative expansion of the data samples based on the original data set through machine learning modeling interpolation prediction to obtain an expanded data set; the data enhancement module is used for constructing an increment information sample data set containing cyclic cycle increment and crack length increment based on the expansion data set; The training modeling module is used for carrying out secondary machine learning modeling based on the increment information sample data set to obtain a fatigue crack growth rate prediction model; The prediction module is used for obtaining a crack length value and a confidence interval thereof at the later stage of the test of the fatigue crack growth of the sample by utilizing the fatigue crack growth rate prediction model and through multipath prediction weighted average, and obtaining a complete fatigue crack growth curve of the sample.

Description

Fatigue crack growth prediction method and system based on incremental information machine learning modeling Technical Field The invention belongs to the technical field of fatigue crack growth prediction, and particularly relates to a fatigue crack growth prediction method and system based on incremental information machine learning modeling. Background Fatigue of metal materials and sudden fracture of components induced by the fatigue are typical failure modes in the engineering field, and serious economic loss and personnel injury are easily caused. Therefore, fatigue properties of materials such as fatigue strength, fatigue life, etc. have been an important basis for reliable design and life assessment of mechanical components in major engineering and critical equipment. Conventional fatigue performance acquisition typically requires a large number of experimental samples, a lengthy test period. The data technology represented by machine learning has the advantages of fitting on complex nonlinear relations, brings new thought and method for optimizing fatigue test experiments and accelerating fatigue performance evaluation, further remarkably improves the material fatigue performance acquisition efficiency, and effectively reduces the fatigue test cost. For example, chinese patent CN117421982a discloses a single-crystal metal material fatigue life prediction method based on machine learning, which is used for predicting fatigue life, reducing experimental cost and improving material use efficiency, chinese patent CN119580902a discloses a material fatigue prediction method based on physical guidance neural network and a system thereof, thereby improving material fatigue prediction precision, chinese patent CN120449668A discloses a metal fatigue life prediction method based on deep learning small sample increment iteration training based on self-attention mechanism and a system thereof, which are used for rapidly obtaining the fatigue limit of the material. Chinese patent CN120524164A discloses a fatigue life assessment method and a system for a data coupling actuating mechanism, which can realize multidimensional data coupling analysis, accurately assess fatigue life and improve assessment accuracy and maintenance efficiency, chinese patent CN120429744A discloses a multi-scale fatigue crack growth prediction method and a system based on a physical information neural network, solves the problem of accurate prediction of multi-scale fatigue crack growth behaviors of metal materials, and Chinese patent CN120509183A discloses a data processing method for fatigue crack growth rate curve piecewise fitting, and accurate fatigue crack growth rate is obtained to provide reliable basis for material engineering application. However, the existing method lacks attention to the crack extension process in the fatigue test, and the fatigue crack extension length still does not have an effective prediction method, so that the acceleration effect of the fatigue test is restricted, the accuracy of fatigue crack extension rule prediction is insufficient, and complex working conditions and data characteristics are difficult to adapt. Disclosure of Invention In order to solve the problems of large prediction deviation and low test efficiency caused by small sample modeling in the existing fatigue crack growth prediction, the invention provides a fatigue crack growth prediction method and a system based on incremental information machine learning modeling, and the efficient and accurate prediction of a fatigue crack growth curve is realized through interpolation prediction iteration data expansion and incremental information-based data enhancement and machine learning modeling. In order to achieve the technical effects, the technical scheme of the invention is as follows: The invention provides a fatigue crack growth prediction method and a system based on incremental information machine learning modeling, comprising the following steps: S1, collecting early data samples of a fatigue crack propagation test of a sample, and taking the early data samples as an original data set; S2, based on an original data set, performing iterative expansion of a data sample through machine learning modeling interpolation prediction to obtain an expanded data set; S3, constructing an increment information sample data set containing cyclic cycle increment and crack length increment based on the expansion data set; S4, performing secondary machine learning modeling based on the increment information sample data set to obtain a fatigue crack growth rate prediction model; S5, obtaining a crack length value and a confidence interval thereof at the later stage of the test of the fatigue crack growth of the sample by utilizing a fatigue crack growth rate prediction model and through multipath prediction weighted average, and obtaining a complete fatigue crack growth curve of the sample. Further, the step of collecting early data samples of the fatigu