Turbine blade fatigue performance influence factor analysis method and system and medium

Abstract

The invention provides a turbine blade fatigue performance influence factor analysis method and system and a medium, and belongs to the technical field of turbine blade performance prediction.The turbine blade fatigue performance influence factor analysis method includes the steps that a small number of training samples are generated, and a back propagation neural network (BP neural network) describing the relation between the fatigue life of a turbine blade and input variables is constructed; a training sample is added to the sequence, and the BP neural network is updated until the calculated failure probability is converged; based on a failure sample obtained in the process of solving the failure probability, estimating conditional failure probability estimation values at different sample points by using a Bayesian inference theory; and obtaining a fatigue reliability sensitivity estimation value of each input variable according to an average difference between the fatigue failure probability estimation value of the turbine blade and the condition failure probability estimation value of each failure sample point. The method solves the problems that when an existing agent model method is used for solving the reliability and sensitivity of the turbine blade, the sample requirement is large, efficiency is low, and the method cannot be suitable for a high-dimensional problem.

Description

Technical Field

[0001] This invention belongs to the field of turbine blade performance prediction technology, specifically relating to a method, system, and medium for analyzing factors affecting the fatigue performance of turbine blades. Background Technology

[0002] Turbine blades, as core components of aero-engines and gas turbines, operate under complex environments of high temperature, high pressure, centrifugal loads, and alternating stress for extended periods. Their fatigue failure directly impacts the safety and reliability of the equipment. Therefore, it is crucial to analyze the fatigue reliability sensitivity of turbine blades in a timely manner to address fatigue failure issues. Traditional methods for solving turbine blade fatigue failure problems employ Monte Carlo simulation and first- / second-order reliability methods. Monte Carlo simulation assesses failure probabilities through random sampling, but requires thousands of finite element analyses, resulting in extremely high computational costs. First- / second-order reliability methods approximate the limit state function through Taylor expansion, but their accuracy is insufficient for high-dimensional nonlinear problems.

[0003] In recent years, machine learning methods have been introduced to replace traditional methods. These primarily involve surrogate models that use training data to establish a mapping between input variables and fatigue life, replacing expensive physical simulations. However, existing machine learning methods require massive amounts of data to simulate the relationship between input variables and fatigue life. Especially when dealing with high-dimensional input variables, the computational cost increases dramatically, making it difficult to meet engineering efficiency requirements. Furthermore, the relationship between turbine blade fatigue life and input variables is highly nonlinear, and traditional models or statistical methods struggle to accurately fit this complex mapping. This results in insufficient accuracy in calculating failure probability and sensitivity, failing to provide reliable support for design optimization. Summary of the Invention

[0004] To address the challenge of establishing a high-fidelity mapping between multiple variables and fatigue life of turbine blades using a small number of samples, while reducing the number of proxy model calls and improving the efficiency of global sensitivity analysis, this invention provides a method for analyzing the influencing factors of turbine blade fatigue performance.

[0005] To achieve the above objectives, the present invention provides the following technical solution: A method for analyzing factors affecting the fatigue performance of turbine blades, including: Multiple structural parameters, material performance parameters, and load parameters of the turbine blade are obtained to form multidimensional input variables. A sample pool with a capacity of N is generated based on the probability density function of each input variable and its corresponding parameters. A portion of samples are randomly selected from the sample pool to form an initial training set. The mapping relationship between the input variables and the fatigue life of the turbine blades is learned through the initial training set, thus completing the initial training of the BP neural network. The fatigue life prediction value corresponding to the sample in the sample pool that was not included in the initial training set is calculated using the initially trained BP neural network. Based on the fatigue life prediction value and life threshold corresponding to the sample that was not included in the initial training set, the sample that is closest to the failure surface among the samples that was not included in the initial training set is added to the initial training set, and the BP neural network is trained again. Repeat the steps of supplementing samples and updating the network multiple times until the convergence condition is met, and a converged BP neural network is obtained. All samples in the sample pool are input into a convergent BP neural network, which outputs the fatigue life prediction value corresponding to each sample. The influence of each input variable on the fatigue performance of the turbine blade is determined by the fatigue life prediction value corresponding to each sample.

[0006] Preferably, determining the influence of each input variable on the fatigue performance of the turbine blade by using the predicted fatigue life values ​​corresponding to each sample specifically includes the following steps: The fatigue life prediction value corresponding to each sample is compared with the life threshold. Failure samples with fatigue life prediction values ​​lower than the life threshold are selected, and the number of failure samples is counted. The fatigue failure probability estimate of the turbine blade is calculated based on the total number of samples and the number of failed samples in the sample pool. Based on failure sample information, the conditional failure probability estimate at a given sample point is estimated using a specified likelihood function and Bayesian inference theory; the given sample point is a set of samples randomly generated by the probability density function of the input variable. Based on the estimated fatigue failure probability of the turbine blade and the estimated conditional failure probability at each failure sample point, the fatigue reliability sensitivity estimate of each input variable is calculated. The influence of each input variable on the fatigue performance of the turbine blade is quantified by the fatigue reliability sensitivity estimate.

[0007] Preferably, the step of repeatedly iterating to supplement samples and update the network is performed multiple times until the convergence condition is met, resulting in a converged BP neural network, specifically: Each time samples are added to the training set and the BP neural network is updated, the fatigue life prediction value corresponding to each sample is calculated once using the updated network. Then the fatigue failure probability estimate of the turbine blade is calculated again. At least 5 updates and calculations are performed continuously to obtain at least 5 fatigue failure probability estimates of the turbine blade. Calculate the mean and standard deviation of the fatigue failure probability estimates for at least five turbine blades, and obtain the final coefficient of variation based on the ratio of the mean and standard deviation. The converged BP neural network is obtained when the coefficient of variation of the estimated failure probability of five consecutive failures is less than a given threshold.

[0008] Preferably, the formula for calculating the fatigue failure probability estimate of the turbine blade is: ; In the formula, This is an estimate of the fatigue failure probability of the turbine blade. This represents the number of failed samples. This represents the total number of samples in the sample pool.

[0009] Preferably, the formula for calculating the conditional failure probability estimate is as follows: ; In the formula, This represents the estimated probability of conditional failure. This represents the number of failed samples. For the sample The One portion, For the first One failure sample The One component; Represents the likelihood function. Indicates input variables The probability density function.

[0010] Preferably, the formula for calculating the fatigue reliability sensitivity estimate is as follows: ; In the formula, This is the fatigue reliability sensitivity estimate.

[0011] Preferably, the failure surface is specifically defined as: when the fatigue life prediction value obtained by mapping the input variable through a BP neural network is exactly equal to the set fatigue life threshold, all sample points that satisfy the correspondence between the input variable and the life threshold together constitute a high-dimensional spatial boundary, and the high-dimensional spatial boundary is the failure surface.

[0012] Preferably, the structural parameters include the position and radius of the air mold hole, the material performance parameters include transverse Young's modulus, longitudinal Young's modulus, transverse Poisson's ratio, longitudinal Poisson's ratio, transverse shear modulus, longitudinal shear modulus, and density, and the load parameters include the maximum speed of the main cycle, the maximum speed of the secondary cycle, and the minimum speed of the secondary cycle.

[0013] This invention also proposes a system for analyzing factors affecting the fatigue performance of turbine blades, comprising: The sample generation module is used to obtain multiple structural parameters, material performance parameters and load parameters of the turbine blade, which constitute multidimensional input variables. Based on the probability density function of each input variable and the corresponding parameters, a sample pool with a capacity of N is generated. The initial training module is used to randomly select a portion of samples from the sample pool to form an initial training set. The initial training set is used to learn the mapping relationship between input variables and turbine blade fatigue life, thus completing the initial training of the BP neural network. The model iterative training module is used to calculate the fatigue life prediction value corresponding to the samples not included in the initial training set in the sample pool using the initially trained BP neural network; based on the fatigue life prediction value and life threshold corresponding to the samples not included in the initial training set, the sample closest to the failure surface among the samples not included in the initial training set is added to the initial training set, and the BP neural network is trained again. Repeat the steps of supplementing samples and updating the network multiple times until the convergence condition is met, and a converged BP neural network is obtained. The analysis module is used to input all samples in the sample pool into a converged BP neural network and output the fatigue life prediction value corresponding to each sample; the influence of each input variable on the fatigue performance of the turbine blade is determined by the fatigue life prediction value corresponding to each sample.

[0014] The present invention also provides a computer-readable storage medium storing a computer program, which, when loaded by a processor, is capable of executing any of the steps in the method for analyzing the influencing factors of turbine blade fatigue performance.

[0015] The method for analyzing factors affecting the fatigue performance of turbine blades provided by this invention has the following beneficial effects: This invention acquires multiple structural parameters, material performance parameters, and load parameters of turbine blades to form multidimensional input variables. Based on the probability density function of each input variable and its corresponding parameters, a sample pool of size N is generated, expanding the initial parameters and reducing the number of initial samples, eliminating the need for massive sample sets. Furthermore, by prioritizing the addition of critical samples most crucial for calculating failure probability, redundant samples are reduced, thereby rapidly improving the fitting accuracy of the BP neural network for the mapping relationship near the failure surface with a small sample base. An effective model can be built with only a small initial training sample and targeted sample addition, significantly improving computational efficiency in high-dimensional input variable scenarios, meeting the efficiency requirements of engineering applications, and reducing sample requirements and computational costs. Simultaneously, the convergent BP neural network can accurately fit the strong nonlinear relationship between input variables and fatigue life. Subsequent input of all samples from the sample pool into the network can output high-precision fatigue life predictions, providing a reliable data foundation for determining the impact of each input variable on fatigue performance. Based on the fatigue life prediction values ​​of each sample output by the convergent BP neural network, the influence of each input variable on the fatigue performance of turbine blades can be further quantified, solving the problem that the traditional model is unreliable in judging the degree of influence due to insufficient fitting accuracy, and providing an effective basis for turbine blade design optimization. Attached Figure Description

[0016] To more clearly illustrate the embodiments and design schemes of the present invention, the accompanying drawings required for this embodiment will be briefly described below. The drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 This is a flowchart of the method for analyzing the factors affecting the fatigue performance of turbine blades proposed in this invention; Figure 2 This is a schematic diagram of the turbine blade simulation structure used in an embodiment of the present invention; Figure 3 This is a stress distribution cloud diagram of the turbine blade structure used in an embodiment of the present invention at maximum speed; wherein, Figure 3 (a) is the overall stress distribution cloud map. Figure 3 (b) is a magnified view of the part outlined in (a); Figure 4 The results are the reliability sensitivity analysis results of the turbine blades used in the embodiments of the present invention. Detailed Implementation

[0018] To enable those skilled in the art to better understand and implement the technical solutions of the present invention, the present invention will be described in detail below with reference to the accompanying drawings and specific embodiments. The following embodiments are only used to more clearly illustrate the technical solutions of the present invention and should not be construed as limiting the scope of protection of the present invention.

[0019] Example 1 Bayesian inference methods can quantify uncertainty through prior and posterior distributions, enabling the extraction of conditional probability features. Therefore, this invention proposes a novel method that integrates the nonlinear fitting capabilities of neural networks with the probabilistic advantages of Bayesian inference to achieve high-precision and high-efficiency fatigue reliability sensitivity prediction of turbine blades, providing effective support for the reliability design and optimization of turbine blades.

[0020] Based on this, the present invention provides a method for analyzing factors affecting the fatigue performance of turbine blades, specifically a method for predicting the fatigue reliability sensitivity of turbine blades based on neural networks and Bayesian inference, such as... Figure 1 As shown, the prediction method includes the following steps: Step S1: Obtain multiple structural parameters, material property parameters, and load parameters of the turbine blade to form multi-dimensional input variables. Generate a sample pool to predict the fatigue reliability of the turbine blade based on the probability density function of each input variable and its corresponding parameters. ,in Represents the capacity of the sample pool. Among them, the structural parameters include the position and radius of the air film holes, the material property parameters include the transverse Young's modulus, longitudinal Young's modulus, transverse Poisson's ratio, longitudinal Poisson's ratio, transverse shear modulus, longitudinal shear modulus, and density, and the load parameters include the maximum rotational speed of the main cycle, the maximum rotational speed of the secondary cycle, and the minimum rotational speed of the secondary cycle.

[0021] Step S2: Randomly select a small number of samples from the sample pool to form an initial training set , and construct a BP neural network to describe the relationship between the fatigue life of the turbine blade and the input variables . Learn the mapping relationship between the input variables and the fatigue life of the turbine blade through the initial training set, and complete the initial training of the BP neural network.

[0022] Step S3: Use the initially trained BP neural network to calculate the fatigue life prediction values corresponding to the samples in the sample pool that are not included in the initial training set; according to the fatigue life prediction values corresponding to the samples not included in the initial training set and the life threshold, add the samples closest to the failure surface among the samples not included in the initial training set to the initial training set, and continue to train the BP neural network.

[0023] Specifically, according to the life threshold in the fatigue reliability analysis of the turbine blade [[ID=IM=19]]select the sample closest to the failure surface to add to the initial training set and reconstruct the BP neural network .

[0024] ; In the formula, represents the sample newly added to the training set .

[0025] In this embodiment, the failure surface is a decision boundary established based on the fatigue life threshold of the turbine blade: when the fatigue life prediction value obtained by mapping the input variables (structural parameters, material property parameters, load parameters) through the BP neural network is exactly equal to the set fatigue life threshold, all the sample points satisfying the corresponding relationship between the input variable - life threshold together constitute the high-dimensional space boundary, which is the failure surface. The failure surface is the basis for screening the samples closest to the failure surface and the failure samples. When screening the failure samples: if the fatigue life prediction value of a certain sample < T0 (i.e., the sample point is on the "failure side" of the failure surface), then this sample is defined as a failure sample; when screening the samples closest to the failure surface, calculate the deviation between the fatigue life prediction value of the sample and the threshold T0 (such as ∣ -T0∣), the sample with the smallest deviation is the sample closest to the failure surface. This type of sample is the most critical for improving the fitting accuracy of the BP neural network near the failure boundary.

[0026] Step S4: Repeat step S3 at least 5 times, and count the coefficient of variation of the failure probability estimates obtained by the BP neural network in the next 5 times, until the coefficient of variation of the fatigue failure probability estimates of the turbine blade in the next 5 times is less than a given threshold, and a converged BP neural network is obtained.

[0027] If the coefficient of variation is less than a given threshold, then the statistics satisfy... Number of failure samples Specifically, the process involves inputting all samples from the sample pool into a converged backpropagation (BP) neural network, which outputs the predicted fatigue life value for each sample. The predicted fatigue life value for each sample is then compared with a life threshold. Failure samples whose predicted fatigue life value is lower than the life threshold are selected, and the number of failure samples is counted.

[0028] Based on the total number of samples and the number of failed samples in the sample pool, the fatigue failure probability estimate of the turbine blade is calculated using the following formula: ; In the formula, The predicted value representing the probability of failure. N This represents the total number of samples.

[0029] If the coefficient of variation is greater than the given threshold, return to step S3.

[0030] Specifically, the coefficient of variation is obtained through the following method: Each time samples are added to the training set and the BP neural network is updated, the fatigue life prediction value corresponding to each sample is calculated once using the updated network. Then, the fatigue failure probability estimate of the turbine blade is calculated again. This process is repeated at least 5 times to obtain the fatigue failure probability estimate of at least 5 turbine blades.

[0031] Calculate the mean and standard deviation of the fatigue failure probability estimates for at least five turbine blades, and obtain the final coefficient of variation based on the ratio of the mean and standard deviation.

[0032] Step S5: Based on the failure sample information, estimate the conditional failure probability at a given sample point using a specified likelihood function and Bayesian inference theory.

[0033] Specifically, the failure sample obtained in step S4 is denoted as... Using these samples, estimate a given sample according to the following formula. The following is an estimate of the probability of failure under certain conditions: ; In the formula, This represents the number of failed samples. The likelihood function is defined as follows: , With a mean of 0 and a standard deviation of The probability density function of a normal distribution. This represents the estimated probability of conditional failure. Indicates input variables The probability density function.

[0034] Step S6: Based on the estimated fatigue failure probability of the turbine blade and the estimated conditional failure probability at each failure sample point, calculate the estimated fatigue reliability sensitivity of the turbine blade using the following formula. , The dimensions of the input variables are defined. The impact of each input variable on the fatigue performance of the turbine blade is quantified using fatigue reliability sensitivity estimates.

[0035] ; In the formula, For the sample The One portion, Indicates sensitivity The estimated value, This represents the estimated probability of fatigue failure of the turbine blade.

[0036] The following describes in detail, with reference to the accompanying drawings, the turbine blade fatigue reliability sensitivity prediction method based on neural networks and Bayesian inference provided in this disclosure: In step S1, consider as follows Figure 2 The simulated turbine blade structure shown includes three main parts: the blade body, the shroud 3, and the tenon 2. The blade body includes structures such as the blade facet 5, the blade back 1, and the film cooling vent 4. The turbine blade is made of nickel-based superalloy DZ125, with a density of [insert density here]. For the coefficient of linear expansion, ultimate strength, conditional yield strength, elastic modulus, shear modulus, Poisson's ratio and specific heat capacity at different temperatures, please refer to the materials handbook.

[0037] The load on turbine blades is mainly considered to be the centrifugal force generated by the high-speed rotation of the blades, which is closely related to the rotational speed. The blade operating temperature is considered to be... The aerodynamic force at the leaf basin is The aerodynamic force on the underside of the leaf is Table 1 shows the simplified main cycle load and secondary cycle load under a single takeoff-landing scenario.

[0038] Table 1. Simplified main cycle and secondary cycle loads under a single takeoff-landing scenario

[0039] By applying loads such as rotational speed and aerodynamic forces to the turbine blade model, the stress distribution cloud map at the maximum rotational speed is obtained, as shown below. Figure 3 As shown. By Figure 3 As shown in (b), the stress around the mold hole closest to the tenon end in the second row of the turbine blade is the greatest. Therefore, this area is considered a dangerous part of the turbine blade for further analysis.

[0040] After determining the assessment points, the fatigue life of the turbine blades was predicted using the Manson-Coffin model with linear correction to Morrow elastic stress. The fatigue life was determined using the least squares method based on experimental data. The fatigue life model is as follows: ; In the formula, For strain amplitude; The maximum stress; It is the elastic modulus; The fatigue life is defined as the sum of the primary and secondary cycles. Miner's linear cumulative damage theory is used to address the cumulative effect of these cycles.

[0041] The main random factors affecting the combined high- and low-cycle fatigue life of turbine blades under primary and secondary cycles include, but are not limited to, structural geometry, material properties, and loads. For the turbine blade structure, the selected random variables of structural geometry include the positions of the air mold holes in column 1. The position of the air mold holes in the second column and air mold hole radius The selected material property random variables include transverse Young's modulus. Longitudinal Young's modulus Horizontal Poisson's ratio Longitudinal Poisson's ratio Transverse shear modulus and longitudinal shear modulus ,density The selected load-type random variables include the maximum rotational speed of the main loop. Maximum speed of the next cycle and the minimum speed of the next cycle The distribution forms and distribution parameters of each input variable are shown in Table 2.

[0042] Let the random input variables composed of the 13 random input variables in Table 2 be denoted as _____. Then, the combined high- and low-cycle fatigue life of a turbine blade under successive cycles can be expressed as the input variable. function .according to The distribution form and distribution parameters generate A sample pool is constructed from 1 sample. .

[0043] Table 2. Distribution form and distribution parameters of input variables in the fatigue reliability sensitivity analysis of turbine blades.

[0044] In step S2, from the sample pool A training set was formed by randomly selecting 130 samples. Constructing a BP neural network .

[0045] In steps S3 and S4, according to... Select new training samples to add Finally, 19 training samples were added to bring the estimated failure probability to convergence. The obtained estimated failure probability is... The number of failed samples is .

[0046] In steps S5 and S6, using 492 failure samples, the fatigue life reliability sensitivity of the turbine blades can be obtained using the Bayesian inference method, as shown in Table 3. Figure 4 As shown. Figure 4 The vertical axis represents the reliability sensitivity of each input variable for the turbine blades, with the sensitivity values ​​on the vertical axis, which visually shows the order of the sensitivity values.

[0047] Based on the sensitivity analysis of turbine blade fatigue reliability, the following conclusions can be drawn: The influence of each parameter on fatigue reliability varies significantly. The maximum speed of the secondary cycle and the maximum speed of the main cycle rank first and second, respectively, indicating that speed parameters have the most critical impact on fatigue reliability. The location of the film cooling vents, transverse Young's modulus, and vent radius rank third to fifth in the second column, indicating that film cooling structure parameters and material transverse stiffness also have a significant impact on fatigue reliability. Parameters such as density and longitudinal Poisson's ratio have relatively smaller impacts, while the longitudinal Young's modulus and transverse Poisson's ratio have the weakest effects. This ranking result suggests that in the optimization design of turbine blade fatigue reliability, priority should be given to optimizing speed parameters and film cooling structure parameters, followed by improving the transverse mechanical properties of the material, while the requirements for longitudinal mechanical property parameters can be appropriately relaxed. This analysis provides clear guidance for the reliability design and parameter optimization of turbine blades, helping to improve the fatigue life of blades while ensuring performance.

[0048] Table 3. Results of fatigue life reliability sensitivity analysis of turbine blades

[0049] This invention directly quantifies the influence of each input sample on the failure probability through fatigue reliability sensitivity, which can clearly distinguish between key and secondary parameters, providing a clear direction for the reliability design and optimization of turbine blades. It helps to prioritize the adjustment of high-sensitivity parameters while ensuring performance, thereby improving the reliability of blade fatigue life. This invention also solves the shortcomings of existing surrogate model methods in solving turbine blade reliability sensitivity, which are either inefficient or unsuitable for high-dimensional problems.

[0050] Based on the same inventive concept, this invention also proposes a system for analyzing factors affecting the fatigue performance of turbine blades, comprising: The sample generation module is used to obtain multiple structural parameters, material performance parameters, and load parameters of the turbine blade, forming multidimensional input variables, and generating a sample pool with a capacity of N based on the probability density function of each input variable and the corresponding parameters.

[0051] The initial training module is used to randomly select a portion of samples from the sample pool to form an initial training set. The initial training set is used to learn the mapping relationship between input variables and turbine blade fatigue life, thus completing the initial training of the BP neural network.

[0052] The model iterative training module is used to calculate the fatigue life prediction value corresponding to the samples in the sample pool that were not included in the initial training set using the initially trained BP neural network; based on the fatigue life prediction value and life threshold corresponding to the samples not included in the initial training set, the sample closest to the failure surface among the samples not included in the initial training set is added to the initial training set, and the BP neural network is trained again.

[0053] Repeat the steps of supplementing samples and updating the network multiple times until the convergence condition is met, resulting in a converged BP neural network.

[0054] The analysis module is used to input all samples in the sample pool into a converged BP neural network and output the fatigue life prediction value corresponding to each sample; the influence of each input variable on the fatigue performance of the turbine blade is determined by the fatigue life prediction value corresponding to each sample.

[0055] The modules in the turbine blade fatigue performance influencing factors analysis system described above can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in the processor of a computer device in hardware form or independent of it, or stored in the memory of a computer device in software form, so that the processor can call and execute the corresponding operations of each module.

[0056] Furthermore, the present invention also provides a non-transitory computer-readable storage medium containing instructions, on which a computer program is stored. For example, a memory containing instructions that can be executed by a processor of a computer device to perform the above-described method. For example, the non-transitory computer-readable storage medium may be a ROM, random access memory (RAM), CD-ROM, magnetic tape, floppy disk, and optical data storage device, etc. When the computer program is executed by the processor, it can implement the steps in the embodiment of the method for analyzing the influencing factors of turbine blade fatigue performance. Specific implementation methods can be found in the method embodiments, which will not be repeated here.

[0057] Those skilled in the art will understand that embodiments of the present invention can provide methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0058] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, as well as combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0059] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0060] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1The steps of the function specified in one or more boxes.

[0061] It should be noted that the specific embodiments described above enable those skilled in the art to more fully understand the present invention, but do not limit the present invention in any way. Therefore, although the present invention has been described in detail in this specification and embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the present invention; and all technical solutions and improvements that do not depart from the spirit and scope of the present invention are covered within the protection scope of the present invention patent. No reference numerals in the claims should be construed as limiting the scope of the claims. Any simple variations or equivalent substitutions of technical solutions that can be readily obtained by those skilled in the art within the scope of the technology disclosed in the present invention are within the protection scope of the present invention.

Claims

1. A method for analyzing factors affecting the fatigue performance of turbine blades, characterized in that, include: Multiple structural parameters, material performance parameters, and load parameters of the turbine blade are obtained to form multidimensional input variables. A sample pool with a capacity of N is generated based on the probability density function of each input variable and its corresponding parameters. A portion of samples are randomly selected from the sample pool to form an initial training set. The mapping relationship between the input variables and the fatigue life of the turbine blades is learned through the initial training set, thus completing the initial training of the BP neural network. The fatigue life prediction value corresponding to the sample in the sample pool that was not included in the initial training set is calculated using the initially trained BP neural network. Based on the fatigue life prediction value and life threshold corresponding to the samples not included in the initial training set, the sample closest to the failure surface among the samples not included in the initial training set is added to the initial training set, and the BP neural network is trained again. Repeat the steps of supplementing samples and updating the network multiple times until the convergence condition is met, and a converged BP neural network is obtained. Input all samples in the sample pool into a converged BP neural network and output the fatigue life prediction value corresponding to each sample. The influence of each input variable on the fatigue performance of the turbine blade is determined by the predicted fatigue life values ​​corresponding to each sample.

2. The method for analyzing factors affecting the fatigue performance of turbine blades according to claim 1, characterized in that, The determination of the influence of each input variable on the fatigue performance of the turbine blade by using the fatigue life prediction values ​​corresponding to each sample specifically includes the following steps: The fatigue life prediction value corresponding to each sample is compared with the life threshold. Failure samples with fatigue life prediction values ​​lower than the life threshold are selected, and the number of failure samples is counted. The fatigue failure probability estimate of the turbine blade is calculated based on the total number of samples and the number of failed samples in the sample pool. Based on failure sample information, the conditional failure probability estimate at a given sample point is estimated using a specified likelihood function and Bayesian inference theory; the given sample point is a set of samples randomly generated by the probability density function of the input variable. Based on the estimated fatigue failure probability of the turbine blade and the estimated conditional failure probability at each failure sample point, the fatigue reliability sensitivity estimate of each input variable is calculated. The influence of each input variable on the fatigue performance of the turbine blade is quantified by the fatigue reliability sensitivity estimate.

3. The method for analyzing factors affecting the fatigue performance of turbine blades according to claim 2, characterized in that, The repeated iterative steps of supplementing samples and updating the network are performed multiple times until the convergence condition is met, resulting in a converged BP neural network. Specifically: Each time samples are added to the training set and the BP neural network is updated, the fatigue life prediction value corresponding to each sample is calculated once using the updated network. Then the fatigue failure probability estimate of the turbine blade is calculated again. At least 5 updates and calculations are performed continuously to obtain at least 5 fatigue failure probability estimates of the turbine blade. Calculate the mean and standard deviation of the fatigue failure probability estimates for at least five turbine blades, and obtain the final coefficient of variation based on the ratio of the mean and standard deviation. The converged BP neural network is obtained when the coefficient of variation of the estimated failure probability of five consecutive failures is less than a given threshold.

4. The method for analyzing factors affecting the fatigue performance of turbine blades according to claim 2, characterized in that, The formula for calculating the fatigue failure probability estimate of the turbine blade is as follows: ； In the formula, This is an estimate of the fatigue failure probability of the turbine blade. This represents the number of failed samples. This represents the total number of samples in the sample pool.

5. The method for analyzing factors affecting the fatigue performance of turbine blades according to claim 4, characterized in that, The formula for calculating the estimated probability of conditional failure is as follows: ； In the formula, This represents the estimated probability of conditional failure. This represents the number of failed samples. For the sample The One portion, For the first One failure sample The One component; Represents the likelihood function. Indicates input variables The probability density function.

6. The method for analyzing factors affecting the fatigue performance of turbine blades according to claim 5, characterized in that, The formula for calculating the fatigue reliability sensitivity estimate is as follows: ； In the formula, This is the fatigue reliability sensitivity estimate.

7. The method for analyzing factors affecting the fatigue performance of turbine blades according to claim 1, characterized in that, The failure surface is specifically defined as follows: when the fatigue life prediction value obtained by mapping the input variable through a BP neural network is exactly equal to the set fatigue life threshold, all sample points that satisfy the correspondence between the input variable and the life threshold together form a high-dimensional spatial boundary, which is the failure surface.

8. The method for analyzing factors affecting the fatigue performance of turbine blades according to claim 1, characterized in that, The structural parameters include the position and radius of the air mold holes; the material property parameters include transverse Young's modulus, longitudinal Young's modulus, transverse Poisson's ratio, longitudinal Poisson's ratio, transverse shear modulus, longitudinal shear modulus, and density; and the load parameters include the maximum speed of the main cycle, the maximum speed of the secondary cycle, and the minimum speed of the secondary cycle.

9. A system for analyzing factors affecting the fatigue performance of turbine blades, characterized in that, include: The sample generation module is used to obtain multiple structural parameters, material performance parameters and load parameters of the turbine blade, which constitute multidimensional input variables. Based on the probability density function of each input variable and the corresponding parameters, a sample pool with a capacity of N is generated. The initial training module is used to randomly select a portion of samples from the sample pool to form an initial training set. The initial training set is used to learn the mapping relationship between input variables and turbine blade fatigue life, thus completing the initial training of the BP neural network. The model iterative training module is used to calculate the fatigue life prediction value of samples in the sample pool that were not included in the initial training set using the initially trained BP neural network. Based on the fatigue life prediction value and life threshold corresponding to the samples not included in the initial training set, the sample closest to the failure surface among the samples not included in the initial training set is added to the initial training set, and the BP neural network is trained again. Repeat the steps of supplementing samples and updating the network multiple times until the convergence condition is met, and a converged BP neural network is obtained. The analysis module is used to input all samples in the sample pool into a converged BP neural network and output the fatigue life prediction value corresponding to each sample. The influence of each input variable on the fatigue performance of the turbine blade is determined by the predicted fatigue life values ​​corresponding to each sample.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is loaded by the processor, it is able to perform the steps of the method according to any one of claims 1 to 7.

Images