A method, system, equipment and medium for analyzing the small failure probability of turbine shaft fatigue life based on a hierarchical surrogate model

Abstract

The present invention relates to the technical field of reliability analysis of complex structures, and specifically to a method, system, device and medium for analyzing the small failure probability of turbine shaft fatigue life based on a hierarchical surrogate model. The method comprises the following steps: S1, constructing a large-capacity candidate sample pool; S2, selecting a small number of samples to construct a Kriging surrogate model; using hypersphere segmentation to generate several small-scale candidate sample pools; S3, training the Kriging model in each small-scale candidate sample pool in turn to determine the number of failure samples in each small-scale candidate sample pool; S4, accumulating the number of failure samples in all small-scale candidate sample pools to obtain the total number of failure samples. N F , and then calculates an estimated value for the turbine blade fatigue life failure probability. This invention uses a large-capacity sample pool segmented by hyperspheres and a Kriging surrogate model to sequentially identify and accumulate failure samples, addressing the problem of poor or even non-convergence in existing surrogate model methods when solving problems with small failure probabilities.

Description

Technical Field

[0001] The present invention relates to the technical field of complex structure reliability analysis, and in particular to a method, system, equipment and medium for analyzing the small failure probability of turbine shaft fatigue life based on a hierarchical proxy model. Background Art

[0002] For core rotating components of aircraft engines such as turbine shafts, the design and manufacturing process must ensure that they can withstand various mechanical and thermal challenges in extreme working environments while maintaining long-term stable operation, thereby ensuring the reliability and safety of the entire aircraft engine system.

[0003] Due to the complex and ever-changing operating environment of turbine shafts, reliability analysis and design of turbine shafts face the challenge of estimating high reliability, specifically, solving for small failure probabilities. This often requires a large pool of candidate samples to ensure accurate failure probability estimation. Training surrogate models within this large pool of candidate samples, however, can lead to time-consuming training and difficulty in convergence, thus compromising the efficiency of reliability analysis. Summary of the Invention

[0004] In response to the problem in the prior art that a large-capacity alternative sample pool will affect the efficiency of reliability analysis due to time-consuming training and difficulty in convergence when solving small failure probability problems, the present invention provides a method, system, equipment and medium for analyzing the small failure probability of turbine shaft fatigue life based on a hierarchical proxy model.

[0005] The present invention is achieved through the following technical solutions:

[0006] A method for analyzing the small failure probability of turbine shaft fatigue life based on a hierarchical surrogate model includes the following steps:

[0007] S1, based on the distribution information of the input variables, a random sampling method is used to construct a large-capacity candidate sample pool with low failure probability of the turbine shaft fatigue life. ;

[0008] S2, from a large-capacity alternative sample pool Select a small number of samples to build a Kriging surrogate model;

[0009] Using a hypersphere to pool large-capacity alternative samples Split and generate several small-scale candidate sample pools ,in, k =1, 2, ..., , represents the total number of small-scale alternative sample pools;

[0010] S3, in each small-scale candidate sample pool Train the Kriging model to identify each small-scale candidate sample pool The number of failed samples;

[0011] S4, for all small-scale candidate sample pools The total number of failure samples is obtained by adding up the number of failure samples in N F , and then calculate the estimated probability of failure of the turbine blade fatigue life .

[0012] Preferably, in S2, the sample size of the initial Kriging surrogate model is set to the random vector dimension ten times.

[0013] Preferably, in S2, each small-scale candidate sample pool The capacity is equal to that of to between.

[0014] Preferably, in S2, a hypersphere is used to select a large sample pool. Split and generate several small-scale candidate sample pools The specific process is as follows: First, the large capacity candidate sample pool The samples in the standard normalization are normalized, and the Euclidean distance between each sample and the coordinate origin is calculated; then, the large-capacity candidate sample pool is sorted in descending order of the Euclidean distance. Sort all samples in the arrive The samples between Samples in

[0015] For the first hypersphere, that is, When the Euclidean distance to the origin is the smallest, the As the sample within the hypersphere; for the second hypersphere, that is, When the Euclidean distance to the origin is the smallest, the to As the sample within the hypersphere; and so on, we can get all samples within the hypersphere, and use this to construct the corresponding sample pool .

[0016] Preferably, in S3, in each small-scale candidate sample pool, The specific process of training the Kriging model is as follows: First, use the Kriging proxy model as the initial model to identify The failure samples in the pool are selected from the candidate samples using the U learning function New training samples are continuously selected and added to the training set until the sample pool The U learning function values ​​of all samples in are greater than or equal to 2;

[0017] For the Alternative sample pools Identification of failed samples in the candidate sample pool The converged Kriging model obtained after training is used as the initial model, and the candidate sample pool is selected according to the U learning function. New training samples are continuously selected and added to the training set until the sample pool The U learning function values ​​of all samples in are greater than or equal to 2.

[0018] Preferably, in S4, the total number of failed samples N F The expression is:

[0019]

[0020] Where, .

[0021] Preferably, an estimate of the probability of failure of turbine blade fatigue life The calculation formula is:

[0022] .

[0023] A system for analyzing the probability of small failure of turbine shaft fatigue life based on a hierarchical surrogate model includes a data acquisition module, a data preprocessing module, a model building module, a model training module, and a data processing module. The data acquisition module is used to construct a large-capacity candidate sample pool with a small probability of small failure of turbine shaft fatigue life. The data preprocessing module is used to use the hypersphere to analyze the large-capacity candidate sample pool. Split and generate several small-scale candidate sample pools , the model building module is used to select from a large pool of candidate samples A small number of samples are selected to build a Kriging surrogate model. The model training module is used to sequentially select samples from each small-scale candidate sample pool. Train the Kriging model to identify each small-scale candidate sample pool The number of failed samples in the data processing module is used to calculate the number of failed samples in the small-scale candidate sample pool. The total number of failure samples is obtained by adding up the number of failure samples in N F , and then calculate the estimated probability of failure of the turbine blade fatigue life .

[0024] An electronic device includes a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the steps of the method for analyzing the small failure probability of fatigue life of a turbine shaft based on a hierarchical proxy model are implemented.

[0025] A storage medium stores a computer program, which, when executed by a processor, implements the steps of a method for analyzing the small failure probability of a turbine shaft fatigue life based on a hierarchical proxy model.

[0026] Compared with the prior art, the present invention has the following beneficial effects:

[0027] The present invention uses a hypersphere to analyze the small failure probability of turbine shaft fatigue life based on a layered proxy model. Divide into several small-scale alternative sample pools , then Training and directly using a large pool of candidate samples Training the Kriging surrogate model In contrast, the supersphere can ensure that each small-scale candidate sample pool They all have certain representativeness and diversity, which helps to improve the training efficiency and accuracy of the proxy model. In addition, the number of failure samples used to solve the failure probability is not a large number of samples from a large-capacity candidate sample pool at one time. Identify statistics, but from several small-scale candidate sample pools This makes the training process more efficient and helps improve the accuracy of the model in specific areas. By accumulating the results, the results can gradually approach the actual failure probability while reducing the error of single recognition statistics. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] Figure 1 This is a flow chart of a method for analyzing the small failure probability of a turbine shaft fatigue life based on a hierarchical proxy model according to the present invention;

[0029] Figure 2 This is a schematic diagram of the basic principle of a method for analyzing the small failure probability of turbine shaft fatigue life based on a hierarchical surrogate model according to the present invention;

[0030] Figure 3 is a schematic structural diagram of a turbine shaft in an embodiment;

[0031] Figure 4 is a force diagram of the turbine shaft in the embodiment;

[0032] Figure 5 is the number of failed samples identified in each small-scale candidate sample pool in the embodiment of the present invention. DETAILED DESCRIPTION

[0033] The present invention will be further described in detail below with reference to specific embodiments, which are intended to explain the present invention rather than to limit it.

[0034] The present invention discloses a method for analyzing the small failure probability of turbine shaft fatigue life based on a hierarchical proxy model. Figure 1 , including the following steps:

[0035] S1, based on the distribution information of the input variables, a random sampling method is used to construct a large-capacity candidate sample pool with low failure probability of the turbine shaft fatigue life. , the capacity of the sample pool N Determined by the following formula:

[0036]

[0037] Where, To estimate the probability of turbine shaft fatigue failure, engineering experience or existing data of similar models can be used.

[0038] S2, from a large-capacity alternative sample pool Use existing tools to randomly select a small number of samples to build a Kriging surrogate model ,in, Indicates the factors that affect the fatigue life of the turbine shaft dimensional random vector, represents the first element of the input vector, represents the second element of the input vector, The first n elements, T Represents the transpose symbol.

[0039] Using a hypersphere to pool large-capacity alternative samples Split and generate several small-scale candidate sample pools ,in, k =1, 2, ..., , Indicates the total number of small-scale candidate sample pools, each of which has The capacity is equal to that of to The specific process of segmentation is as follows:

[0040] First, a large-capacity alternative sample pool The samples in the standard normalization are normalized, and the Euclidean distance between each sample and the coordinate origin is calculated; then, the large-capacity candidate sample pool is sorted in descending order of the Euclidean distance. Sort all samples in the arrive The samples between For the first hypersphere, that is, When the Euclidean distance to the origin is the smallest, the As the sample within the hypersphere; for the second hypersphere, that is, When the Euclidean distance to the origin is the smallest, the to As the sample within the hypersphere; and so on, we can get all samples within the hypersphere, and use this to construct the corresponding sample pool .

[0041] S3, in each small-scale candidate sample pool Train the Kriging model to identify each small-scale candidate sample pool The number of failed samples in is denoted as The specific process is as follows:

[0042] First, the Kriging surrogate model established in step S2 is used as the initial model to identify In order to ensure the recognition accuracy of the failure samples in the pool, the U learning function is used to select the candidate samples from the pool. New training samples are continuously selected and added to the training set until the sample pool The U learning function values ​​of all samples in are greater than or equal to 2; Alternative sample pools Identification of failed samples in the candidate sample pool The converged Kriging model obtained after training is used as the initial model, and the candidate sample pool is selected according to the U learning function. New training samples are continuously selected and added to the training set until the sample pool The U learning function values ​​of all samples in are greater than or equal to 2.

[0043] S4, for all small-scale candidate sample pools The total number of failure samples is obtained by adding up the number of failure samples in Then according to Calculate an estimate of the probability of failure during the fatigue life of a turbine blade .

[0044] This paper presents a method for analyzing the small failure probability of turbine shaft fatigue life based on a hierarchical surrogate model. By employing innovative methods such as sample pool segmentation, hierarchical surrogate model training, and a gradual accumulation of failure samples, it effectively addresses the shortcomings of traditional methods in terms of computational resource consumption, training time, and model accuracy. This method provides a new approach and method for turbine shaft fatigue life analysis and has significant practical application value.

[0045] The following describes in detail the hierarchical proxy model method for analyzing the small failure probability of a turbine shaft fatigue life provided by the embodiment of the present disclosure with reference to the accompanying drawings:

[0046] In step 1, consider Figure 3 The turbine shaft structure shown is made of GH4169 material and has an operating temperature of 360°C. During operation, the turbine shaft is subjected to rotational force, axial force, moment of inertia, and working torque, as shown in the following figure. Figure 4 Considering the three mission profiles of "start-maximum-start," "idle-maximum-idle," and "cruise-maximum-cruise," the turbine shaft speed and cycle number for each profile are shown in Table 1. The maximum stress and strain of the turbine shaft structure obtained through finite element analysis under the three operating conditions of "maximum," "idle," and "cruise" are shown in Table 2.

[0047] Table 1. Focus elements of turbine shaft structure input variables and corresponding basic probability distribution

[0048]

[0049] Table 2 Maximum stress and maximum strain of turbine shaft structure input variables under different operating conditions

[0050]

[0051] According to the Manson-Coffin equation, the fatigue life of the turbine shaft under various mission sections can be predicted. Then, the fatigue life of the turbine shaft structure can be calculated using the linear cumulative damage theory. The fatigue life is calculated under the condition of deterministic input variables. However, due to the influence of various uncertain factors, input variables such as material properties and external loads are usually uncertain. Therefore, the fatigue life of the turbine shaft structure is can be considered as an input vector The function of In this embodiment, five input variables are considered, including the inner and outer diameters of the vent hole, the elastic modulus and Poisson's ratio of the material, and the maximum rotational speed. The distribution forms and distribution parameters are shown in Table 3.

[0052] Table 3 Distribution form and distribution parameters of input variables affecting turbine shaft fatigue life

[0053]

[0054] Considering that the fatigue life of the turbine shaft structure is less than the designed fatigue life This failure mode can be used to establish the following function in the turbine shaft fatigue life reliability analysis:

[0055]

[0056] In this embodiment, The corresponding fatigue life failure probability can be expressed as:

[0057]

[0058] Where, represents a probability operator.

[0059] The fatigue life failure probability of the turbine shaft is estimated to be Therefore, according to the distribution form and parameters of each input variable in Table 3, the capacity can be generated as Alternative sample pool .

[0060] In step 2, since the input vector in this embodiment is 5, the candidate sample pool Randomly select 50 samples to build the initial Kriging surrogate model .

[0061] In step 3, by ensuring that the capacity of each small-scale candidate sample pool is ,Will Divided into 500 small-scale alternative sample pools .

[0062] In step 4, the Kriging proxy model is updated in each candidate sample pool in turn. , and identify the failed samples. The number of failed samples in each sample pool is as follows: Figure 5 As shown (the horizontal axis represents the number of small-scale alternative sample pools, and the vertical axis represents the number of failed samples), the final number of training samples required is 436.

[0063] In step 5, the total number of failure samples in all small-scale candidate sample pools is accumulated to obtain 1680, so the turbine shaft fatigue life failure probability can be calculated as .

[0064] The present invention also discloses a system for analyzing the probability of small failure of turbine shaft fatigue life based on a hierarchical proxy model, comprising a data acquisition module, a data preprocessing module, a model building module, a model training module and a data processing module. The data acquisition module is used to construct a large-capacity candidate sample pool of the probability of small failure of turbine shaft fatigue life. The data preprocessing module is used to use the hypersphere to analyze the large-capacity candidate sample pool. Split and generate several small-scale candidate sample pools , the model building module is used to select from a large pool of candidate samples A small number of samples are selected to build a Kriging surrogate model. The model training module is used to sequentially select samples from each small-scale candidate sample pool. Train the Kriging model to identify each small-scale candidate sample pool The number of failed samples in the data processing module is used to calculate the number of failed samples in the small-scale candidate sample pool. The total number of failure samples is obtained by adding up the number of failure samples in N F , and then calculate the estimated probability of failure of the turbine blade fatigue life .

[0065] The present invention also discloses an electronic device, including a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the steps of a method for analyzing the small failure probability of fatigue life of a turbine shaft based on a hierarchical proxy model are implemented.

[0066] The present invention also discloses a storage medium on which a computer program is stored. When the computer program is executed by a processor, the steps of a method for analyzing the small failure probability of fatigue life of a turbine shaft based on a hierarchical proxy model are implemented.

[0067] The above description is merely a preferred embodiment of the present invention and is not intended to impose any limitation on the technical solution of the present invention. Those skilled in the art should understand that, without departing from the spirit and principles of the present invention, the technical solution can also be subjected to several simple modifications and replacements, and these modifications and replacements are also within the scope of protection covered by the claims.

Claims

1. A method for analyzing the small failure probability of turbine shaft fatigue life based on a hierarchical surrogate model, characterized in that: The following steps are involved: S1, based on the distribution information of the input variables, a random sampling method is used to construct a large-capacity candidate sample pool with low failure probability of the turbine shaft fatigue life. ; S2, from a large-capacity alternative sample pool Select a small number of samples to build a Kriging surrogate model; Using a hypersphere to pool large-capacity alternative samples Split and generate several small-scale candidate sample pools ,in, k =1, 2, ..., , represents the total number of small-scale alternative sample pools; Using a hypersphere to pool large-capacity alternative samples Split and generate several small-scale candidate sample pools The specific process is as follows: First, the large capacity candidate sample pool The samples in the standard normalization are normalized, and the Euclidean distance between each sample and the coordinate origin is calculated; then, the large-capacity candidate sample pool is sorted in descending order of the Euclidean distance. Sort all samples in the arrive The samples between Samples in For the first hypersphere, that is When the Euclidean distance to the origin is the smallest, the As the sample within the hypersphere; for the second hypersphere, that is, When the Euclidean distance to the origin is the smallest, the to As the sample within the hypersphere; and so on, we can get all samples within the hypersphere, and use this to construct the corresponding sample pool ; in, is the sample pool capacity, which is determined by the following formula: Where, is the estimated probability of turbine shaft fatigue failure; S3, in each small-scale candidate sample pool Train the Kriging model to identify each small-scale candidate sample pool The number of failed samples; In each small-scale candidate sample pool The specific process of training the Kriging model is as follows: First, use the Kriging proxy model as the initial model to identify The failure samples in the pool are selected from the candidate samples using the U learning function New training samples are continuously selected and added to the training set until the sample pool The U learning function values ​​of all samples in are greater than or equal to 2; For the , Alternative sample pools Identification of failed samples in the candidate sample pool The converged Kriging model obtained after training is used as the initial model, and the candidate sample pool is selected according to the U learning function. New training samples are continuously selected and added to the training set until the sample pool The U learning function values ​​of all samples in are greater than or equal to 2; S4, for all small-scale candidate sample pools The total number of failure samples is obtained by adding up the number of failure samples in N F , and then calculate the estimated probability of failure of the turbine blade fatigue life .

2. The method for analyzing the small failure probability of turbine shaft fatigue life based on a hierarchical proxy model according to claim 1 is characterized in that: In S2, the sample size of the initial Kriging surrogate model is set to the random vector dimension ten times.

3. The method for analyzing the small failure probability of turbine shaft fatigue life based on a hierarchical proxy model according to claim 1 is characterized in that: In S2, each small-scale candidate sample pool The capacity is equal to that of to between.

4. The method for analyzing the small failure probability of turbine shaft fatigue life based on a hierarchical proxy model according to claim 1 is characterized in that: In S4, the total number of failed samples N F The expression is: Where, .

5. The method for analyzing the small failure probability of turbine shaft fatigue life based on a hierarchical surrogate model according to claim 1 is characterized in that: Estimation of Fatigue Life Failure Probability of Turbine Blades The calculation formula is: 。 6. A system for analyzing the small failure probability of a turbine shaft fatigue life based on a hierarchical proxy model according to any one of claims 1 to 5, characterized in that: It includes data acquisition module, data preprocessing module, model building module, model training module and data processing module. The data acquisition module is used to build a large-capacity candidate sample pool with low failure probability of turbine shaft fatigue life. The data preprocessing module is used to use the hypersphere to analyze the large-capacity candidate sample pool. Split and generate several small-scale candidate sample pools , the model building module is used to select from a large pool of candidate samples A small number of samples are selected to build a Kriging surrogate model. The model training module is used to sequentially select samples from each small-scale candidate sample pool. Train the Kriging model to identify each small-scale candidate sample pool The number of failed samples in the data processing module is used to calculate the number of failed samples in the small-scale candidate sample pool. The total number of failure samples is obtained by adding up the number of failure samples in N F , and then calculate the estimated probability of failure of the turbine blade fatigue life .

7. An electronic device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method for analyzing the small failure probability of fatigue life of a turbine shaft based on a hierarchical proxy model according to any one of claims 1 to 5 are implemented.

8. A storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method for analyzing the small failure probability of fatigue life of a turbine shaft based on a hierarchical proxy model according to any one of claims 1 to 5 are implemented.

Images