Helicopter transmission system defect tolerance evaluation method and system

Abstract

The invention belongs to the technical field of aero-engines, and discloses a helicopter transmission system defect tolerance evaluation method and system.The method comprises the steps that a transmission system is divided into a plurality of to-be-evaluated areas based on surface geometric features of a typical structure of the transmission system, and the sampling proportion of each to-be-evaluated area is obtained based on Bayesian probability fusion; inputting the input parameters obtained by sampling into the fatigue life agent model of the structure containing the defect, thereby obtaining the fatigue life distribution of each partition, further obtaining a fatigue life value under the given defect tolerance reliability, and solving the defect tolerance reliability of the typical structure of the helicopter by adopting a series reliability system theory and cooperating with the reliability of all the partitions, so as to obtain the reliability of the typical structure of the helicopter. The method can consider possible defects in all areas of the structure surface, and can be used for scenes with complex uncertainty factors and multivariable interaction influence.

Description

Technical Field

[0001] The invention belongs to the technical field of aero-engines, and in particular relates to a method and system for evaluating defect tolerance of a helicopter transmission system. Background Art

[0002] The helicopter transmission system is one of the core components of the helicopter. Its main function is to transfer the power of the engine to the rotor system to achieve the normal flight of the helicopter. Since the transmission system operates under high load, high speed and complex environment for a long time, various structural defects (such as cracks, corrosion, wear, etc.) are inevitable. These defects will not only affect the normal operation of the transmission system, but may also cause serious safety accidents. Therefore, conducting tolerance assessment research on typical structural defects in the transmission system is of great significance to improving the safety and reliability of helicopters.

[0003] At present, the evaluation methods for helicopter transmission system structural defects mainly include traditional stress analysis, fatigue life prediction and finite element simulation, etc. However, these methods have certain limitations when dealing with complex uncertain factors and multivariate interactions.

[0004] For example, traditional stress analysis methods cannot fully consider uncertainties such as material properties, environmental factors, and workloads; fatigue life prediction methods are highly dependent on defect morphology and stress distribution, and are difficult to adapt to a variety of defect forms; although finite element simulation methods can provide more detailed stress and deformation information, they are computationally intensive, time-consuming, and require high-precision models and a large amount of experimental data support. In addition, the study of fatigue failure problems in defective structures often uses prefabricated defects in dangerous areas, which is called the hotspot method. This method artificially biases the probability sampling problem, making the final structural fatigue life too safe, resulting in the scrapping of the equipment before it completely fails.

[0005] In addition, the existing defect assessment methods are mainly aimed at life assessment methods after defects are generated in exposed structural parts during use. They are life assessment methods based on the determination of defects. However, the actual defects (defect type, defect location, defect size) are randomly scattered. At present, there is a lack of defect tolerance assessment methods for helicopter transmission systems that take reliability into consideration.

[0006] Therefore, it is necessary to propose a method that can reasonably evaluate the structural defect tolerance, so that possible defects in all areas of the structure surface can be considered and can be used in scenarios with complex uncertainties and multivariable interactions. Summary of the invention

[0007] In view of the above problems, the present invention provides a method and system for evaluating defect tolerance of a helicopter transmission system, which adopts the following technical solutions:

[0008] A method for evaluating the defect tolerance of a helicopter transmission system, comprising:

[0009] Based on the geometric features of the surface of the typical structure of the helicopter transmission system, the typical structure is divided into multiple areas to be evaluated;

[0010] According to the stress analysis results of the finite element simulation of the typical structure without defects, a stress probability distribution model for each area to be evaluated is established; according to the fatigue life model and the statistical results of external field defects, an input parameter probability distribution model is established;

[0011] According to the parameters obtained from the stress probability distribution model of the area to be evaluated and the input parameter probability distribution model, a surrogate model sample is constructed, and the fatigue life surrogate model of the area to be evaluated is trained and tested through the surrogate model sample to obtain the trained fatigue life surrogate model;

[0012] Based on Bayesian probability fusion, the sampling ratio of each area to be evaluated is obtained. According to the sampling ratio, the input parameters of the surrogate model are obtained from the input parameter probability distribution model of each area to be evaluated, and the input parameters of the surrogate model are input into the trained fatigue life surrogate model to obtain the fatigue life distribution of each area to be evaluated;

[0013] According to the fatigue life distribution of the area to be evaluated, the fatigue life value under the given defect tolerance reliability of the area to be evaluated is determined; based on the theory of series reliability system, according to the fatigue life values under the given defect tolerance reliability of all areas to be evaluated, the defect tolerance reliability of the typical structure is determined.

[0014] Further, based on the geometric features of the surface of the typical structure of the helicopter transmission system, dividing the typical structure into multiple areas to be evaluated includes the following steps:

[0015] According to the curved surface, splicing and holes on the surface of the typical structure of the transmission system, the surface of the typical structure is divided into multiple areas to be evaluated, where each area to be evaluated is the exposed surface outside the typical structure and there are defects falling in this area to be evaluated in the statistical results of external field defects.

[0016] Further, according to the stress analysis results of the finite element simulation of the typical structure without defects, establishing a stress probability distribution model for each area to be evaluated includes the following steps:

[0017] For the typical structure without defects, through material setting, mesh division and boundary condition loading operations, the finite element simulation results of the typical structure under typical working conditions are obtained, and the stress results of different areas to be evaluated are derived from the finite element simulation results;

[0018] Based on the theory of probability and statistics, a stress distribution histogram of each area to be evaluated is established, and the stress probability distribution model of each area to be evaluated is fitted by gamma distribution.

[0019] Furthermore, the input parameter probability distribution model includes a defect morphology parameter distribution model, a fatigue life model parameter distribution model, and a material parameter distribution model.

[0020] Furthermore, an input parameter probability distribution model is established based on the fatigue life model and the statistical results of external field defects, including the following steps:

[0021] Obtain defect morphology parameters, fatigue life parameters, and material parameters through the statistical results of external field defects, and use the normal distribution to respectively fit the distribution means and variances of the defect morphology parameters, fatigue life parameters, and material parameters to obtain the defect morphology parameter distribution model, the fatigue life model parameter distribution model, and the material parameter distribution model.

[0022] Furthermore, the defect morphology parameters include defect depth and defect bottom radius, and the material parameters include elastic modulus, fatigue ductility coefficient, stress amplitude, and fatigue ductility index.

[0023] Furthermore, construct a surrogate model sample based on the parameters obtained from the stress probability distribution model of the area to be evaluated and the input parameter probability distribution model, including the following steps:

[0024] Sample all the stress probability distribution models of the areas to be evaluated to obtain defect location samples;

[0025] Perform combined sampling on the probability quantization model of the surrogate model input parameters to obtain the surrogate model input parameters, which include defect morphology parameters, fatigue life model parameters, and material parameters;

[0026] Correct the stress of each node in the defect location sample according to the defect morphology parameters and the stress correction factor to obtain the corrected stress of the defect location, and input the maximum corrected stress, fatigue life model parameters, and material parameters into the fatigue life model to obtain fatigue life samples;

[0027] Use the surrogate model input parameters as the input and the fatigue life samples as the output to construct a surrogate model sample.

[0028] Furthermore, train and test the fatigue life surrogate model of the area to be evaluated through the surrogate model sample to obtain the trained fatigue life surrogate model, including the following steps:

[0029] Divide the surrogate model sample into a training set and a test set;

[0030] A fatigue life surrogate model for a defective structure is established based on a genetic algorithm and a Kriging model. Hyperparameters of the fatigue life surrogate model are obtained through a training set, and the accuracy of the fatigue life surrogate model is verified through a test set. If the accuracy meets the requirements, the fatigue life surrogate model is output. If the accuracy does not meet the requirements, the number of samples in the training set is increased and retraining is performed until the accuracy requirements are met, and a trained fatigue life surrogate model is obtained.

[0031] Furthermore, based on Bayesian probability fusion, the sampling ratio of each region to be evaluated is obtained. According to the sampling ratio, input parameters of the surrogate model are obtained from the probability distribution model of the input parameters of each region to be evaluated. The input parameters of the surrogate model are input into the trained fatigue life surrogate model to obtain the fatigue life distribution of each region to be evaluated, including the following steps:

[0032] Obtain the area of the characteristics of each region to be evaluated, and then determine the sampling weight of each region to be evaluated based on the area;

[0033] Use the probability of external field statistical defects appearing on the surface to bias the sampling weight, and determine the sampling quantity of the region to be evaluated based on the biased sampling weight;

[0034] According to the determined sampling quantity of the region to be evaluated, use Latin hypercube to obtain the input parameters of the surrogate model from the probability distribution model of the input parameters of each region to be evaluated, substitute them into the trained fatigue life surrogate model, obtain multiple life distributions and the quantity of the life of each region to be evaluated, and calculate the frequency of each life occurrence to obtain the life distribution histogram of each region to be evaluated.

[0035] Furthermore, according to the fatigue life distribution of the region to be evaluated, determine the fatigue life value under the given defect tolerance reliability of the region to be evaluated; based on the series reliability system theory, according to the fatigue life values under the given defect tolerance reliability of all regions to be evaluated, determine the defect tolerance reliability of the typical structure, including the following steps:

[0036] Use the distribution fitting method to fit the established fatigue life probability distribution histograms of each region to be evaluated into probability density distribution curves, and based on the cumulative integral function, convert the probability density distribution curves into reliability and failure rate curves of fatigue life. Determine the fatigue life value under the given defect tolerance reliability of the region to be evaluated according to the reliability and failure rate curves of fatigue life;

[0037] Based on the collaborative reliability assessment theory of the region to be evaluated, determine the defect tolerance reliability of the overall structure of the helicopter transmission system according to the fatigue life value under the given defect tolerance reliability of the region to be evaluated, and complete the defect tolerance reliability analysis of the typical structure of the transmission system.

[0038] The present invention also provides a helicopter transmission system defect tolerance assessment system, including:

[0039] The region division module is used to divide the typical structure into multiple regions to be evaluated based on the geometric features of the surface of the typical structure of the helicopter transmission system;

[0040] The model establishment module is used to establish a stress probability distribution model for each region to be evaluated according to the stress analysis results of the finite element simulation of the typical structure without defects; establish an input parameter probability distribution model according to the fatigue life model and the statistical results of external field defects;

[0041] The model training module is used to construct a surrogate model sample according to the parameters obtained from the stress probability distribution model and the input parameter probability distribution model of the region to be evaluated, and train and test the fatigue life surrogate model of the region to be evaluated through the surrogate model sample to obtain the trained fatigue life surrogate model;

[0042] The first calculation module is used to obtain the sampling ratio of each region to be evaluated based on Bayesian probability fusion, obtain the surrogate model input parameters from the input parameter probability distribution model of each region to be evaluated according to the sampling ratio, and input the surrogate model input parameters into the trained fatigue life surrogate model to obtain the fatigue life distribution of each region to be evaluated;

[0043] The second calculation module is used to determine the fatigue life value under the given defect tolerance reliability of the region to be evaluated according to the fatigue life distribution of the region to be evaluated; based on the series reliability system theory, determine the defect tolerance reliability of the typical structure according to the fatigue life values under the given defect tolerance reliability of all regions to be evaluated.

[0044] Advantages of the present invention:

[0045] The present invention divides the typical structure of the transmission system into multiple regions to be evaluated based on the surface geometric features of the typical structure, obtains the sampling ratio of each region to be evaluated based on Bayesian probability fusion, and inputs the sampled input parameters into the fatigue life surrogate model of the defective structure, so as to obtain the fatigue life distribution of each partition, and further obtain the fatigue life value under the given defect tolerance reliability, and adopts the series reliability system theory to jointly solve the defect tolerance reliability of the typical structure of the helicopter with the reliability of all partitions. The present invention can consider the defects that may appear in all regions of the structure surface and can be used in scenarios with complex uncertainty factors and multivariable interaction effects.

[0046] Other features and advantages of the present invention will be described in the subsequent specification, and, in part, will be obvious from the specification, or will be understood by implementing the present invention. The objectives and other advantages of the present invention can be realized and obtained by the structures pointed out in the specification and the drawings. Brief Description of the Drawings

[0047] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the drawings in the following description are some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0048] Figure 1 Shows a schematic diagram of the main process of a method for evaluating the defect tolerance of a helicopter transmission system according to an embodiment of the present invention;

[0049] Figure 2 Shows a schematic diagram of the detailed process of a method for evaluating the defect tolerance of a helicopter transmission system according to an embodiment of the present invention;

[0050] Figure 3 Shows a schematic diagram of dividing the area to be evaluated for the tail reduction gearbox according to an embodiment of the present invention;

[0051] Figure 4 Shows a schematic diagram of the finite element node distribution and stress value magnitude of geometric partition 1 of the tail reduction gearbox according to an embodiment of the present invention;

[0052] Figure 5 Shows a schematic diagram of the finite element node distribution and stress value magnitude of geometric partition 2 of the tail reduction gearbox according to an embodiment of the present invention;

[0053] Figure 6 Shows a schematic diagram of the finite element node distribution and stress value magnitude of geometric partition 3 of the tail reduction gearbox according to an embodiment of the present invention;

[0054] Figure 7 Shows a schematic diagram of the finite element node distribution and stress value magnitude of geometric partition 4 of the tail reduction gearbox according to an embodiment of the present invention;

[0055] Figure 8 Shows a histogram of the finite element node distribution of geometric partition 1 of the tail reduction gearbox according to an embodiment of the present invention;

[0056] Figure 9 Shows a histogram of the finite element node distribution of geometric partition 2 of the tail reduction gearbox according to an embodiment of the present invention;

[0057] Figure 10 Shows a histogram of the finite element node distribution of geometric partition 3 of the tail reduction gearbox according to an embodiment of the present invention;

[0058] Figure 11 Shows a histogram of the finite element node distribution of geometric partition 4 of the tail reduction gearbox according to an embodiment of the present invention;

[0059] Figure 12Shows the histogram of the depth and radius outer field statistical results of defects according to an embodiment of the present invention;

[0060] Figure 13 Shows the sampling samples of the defect positions in all regions of the tail reduction casing surface according to an embodiment of the present invention;

[0061] Figure 14 Shows the calculation process of the fatigue life of the defective structure of the tail reduction casing based on the stress correction factor according to an embodiment of the present invention;

[0062] Figure 15 Shows the histogram of the fatigue life sample distribution of geometric partition 1 of the defective tail reduction casing according to an embodiment of the present invention;

[0063] Figure 16 Shows the histogram of the fatigue life sample distribution of geometric partition 2 of the defective tail reduction casing according to an embodiment of the present invention;

[0064] Figure 17 Shows the histogram of the fatigue life sample distribution of geometric partition 3 of the defective tail reduction casing according to an embodiment of the present invention;

[0065] Figure 18 Shows the histogram of the fatigue life sample distribution of geometric partition 4 of the defective tail reduction casing according to an embodiment of the present invention;

[0066] Figure 19 Shows the convergence curve of the GA-Kriging surrogate model of geometric partition 1 of the tail reduction casing according to an embodiment of the present invention;

[0067] Figure 20 Shows the convergence curve of the GA-Kriging surrogate model of geometric partition 2 of the tail reduction casing according to an embodiment of the present invention;

[0068] Figure 21 Shows the convergence curve of the GA-Kriging surrogate model of geometric partition 3 of the tail reduction casing according to an embodiment of the present invention;

[0069] Figure 22 Shows the convergence curve of the GA-Kriging surrogate model of geometric partition 4 of the tail reduction casing according to an embodiment of the present invention;

[0070] Figure 23 Shows the prediction accuracy and error analysis of the fatigue life surrogate model of geometric partition 1 of the tail reduction casing according to an embodiment of the present invention;

[0071] Figure 24 Shows the prediction accuracy and error analysis of the fatigue life surrogate model of geometric partition 2 of the tail reduction casing according to an embodiment of the present invention;

[0072] Figure 25 Shows the prediction accuracy and error analysis of the fatigue life surrogate model for the geometric partition 3 of the tail reduction casing according to an embodiment of the present invention;

[0073] Figure 26 Shows the prediction accuracy and error analysis of the fatigue life surrogate model for the geometric partition 4 of the tail reduction casing according to an embodiment of the present invention;

[0074] Figure 27 Shows the fatigue distribution histogram of the geometric partition 1 of the tail reduction casing according to an embodiment of the present invention;

[0075] Figure 28 Shows the fatigue distribution histogram of the geometric partition 2 of the tail reduction casing according to an embodiment of the present invention;

[0076] Figure 29 Shows the fatigue distribution histogram of the geometric partition 3 of the tail reduction casing according to an embodiment of the present invention;

[0077] Figure 30 Shows the fatigue distribution histogram of the geometric partition 4 of the tail reduction casing according to an embodiment of the present invention;

[0078] Figure 31 Shows the probability density distribution curve of the geometric partition 1 of the tail reduction casing according to an embodiment of the present invention, as well as the schematic diagrams of the reliability and failure rate curves of the fatigue life;

[0079] Figure 32 Shows the probability density distribution curve of the geometric partition 2 of the tail reduction casing according to an embodiment of the present invention, as well as the schematic diagrams of the reliability and failure rate curves of the fatigue life;

[0080] Figure 33 Shows the probability density distribution curve of the geometric partition 3 of the tail reduction casing according to an embodiment of the present invention, as well as the schematic diagrams of the reliability and failure rate curves of the fatigue life;

[0081] Figure 34 Shows the probability density distribution curve of the geometric partition 4 of the tail reduction casing according to an embodiment of the present invention, as well as the schematic diagrams of the reliability and failure rate curves of the fatigue life;

[0082] Figure 35 Shows the structural schematic diagram of a helicopter transmission system defect tolerance assessment system according to an embodiment of the present invention. Specific embodiments

[0083] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Apparently, the described embodiments are some, but not all, of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0084] It should be noted that the terms "first", "second", etc. in this application are used to distinguish similar objects and do not necessarily describe a specific order or sequence. It should be understood that such data can be interchanged under appropriate circumstances so as to implement the embodiments of the present application described herein.

[0085] The embodiments of the present invention provide a method and system for evaluating the defect tolerance of a helicopter transmission system, which solves the problem of collaborative reliability evaluation of typical structural partitions of a helicopter transmission system with defective structures, and establishes a method for evaluating the defect tolerance of a single partition and a method for evaluating the overall defect tolerance, and can also be used for other helicopters.

[0086] As Figure 1 and Figure 2 shown, a method for evaluating the defect tolerance of a helicopter transmission system includes the following steps:

[0087] S1. Based on the geometric features of the surface of the typical structure of the helicopter transmission system, the typical structure is divided into multiple regions to be evaluated, including the following steps:

[0088] According to the curved surface, splicing, and holes on the surface of the typical structure of the transmission system, the surface of the typical structure is divided into multiple regions to be evaluated, which are labeled as partition 1, partition 2,..., partition n respectively. Among them, each region to be evaluated is the exposed surface outside the typical structure, and defects are shown to fall in this region to be evaluated in the statistical results of on-site defects.

[0089] For example, the typical structural parts of the transmission system include flange plates, transmission shafts, and tail reduction casings. As Figure 3 shown, taking the tail reduction casing as an example to illustrate the division of the regions to be evaluated, based on the geometric shape of the tail reduction casing of the transmission system, the regions where defects may occur and areas with high stress concentration are taken as the regions to be evaluated, and based on its geometric features, it is divided into four regions to be evaluated (partition 1-4). Among them, the first side at the bottom of the tail reduction casing is taken as partition 1, the outer side of the perimeter of the inspection hole at the bottom of the tail reduction casing is taken as partition 2, the inner side wall of the inspection hole at the bottom of the tail reduction casing is taken as partition 3, and the second side at the bottom of the tail reduction casing is taken as partition 4. The first side and the second side are two opposite sides, and the inspection hole penetrates the first side and the second side.

[0090] S2. Based on the stress analysis results of the finite element simulation of the defect-free typical structure, establish the stress probability distribution model for each area to be evaluated; establish the probability distribution model of input parameters according to the fatigue life model and the statistical results of external field defects.

[0091] Among them, based on the stress analysis results of the finite element simulation of the defect-free typical structure, establishing the stress probability distribution model for each area to be evaluated includes the following steps:

[0092] S21. For the defect-free typical structure, through material setting, mesh generation, and boundary condition loading operations, obtain the finite element simulation results of the typical structure under typical working conditions, and derive the stress results of different areas to be evaluated from the finite element simulation results.

[0093] Taking the tail reduction casing as an example, through material setting, mesh generation, and boundary condition loading operations, obtain the finite element simulation results of the tail reduction casing under typical working conditions, derive the stress results of different surface areas, and respectively obtain the nodal stress distribution nephograms of four partitions, as Figure 4 、 Figure 5 、 Figure 6 and Figure 7 shown.

[0094] S22. Based on the probability statistics theory, establish the stress distribution histogram for each area to be evaluated, and fit the stress probability distribution model for each area to be evaluated through the gamma distribution. The probability density function of the gamma distribution is defined as:

[0095]

[0096] Among them, Γ(·) is the gamma function, where a is the shape parameter, b is the scale parameter, x is the random variable representing the stress value of the area to be evaluated, and y is the value of the probability density function representing the relative possibility of the specific stress value x appearing. In addition, the cumulative distribution function of the gamma distribution is:

[0097]

[0098] In the formula, t is the integration variable, representing all possible stress values in the range from 0 to x, and the value range of t is [0, x].

[0099] Through the probability density function and cumulative distribution function of the gamma distribution, the values of the fitting parameters a and b of the stress of different areas to be evaluated can be obtained, so as to obtain the probability distribution model of the area to be evaluated.

[0100] Taking the tail reduction casing as an example, in order to establish the defect location of each area, based on the probability statistics theory, establish the stress distribution histogram of each area to be evaluated, and then fit the probability distribution model of each partition through the gamma distribution. The stress distribution histogram of each area to be evaluated is as Figure 8, Figure 9 , Figure 10 and Figure 11 as shown. Using the distribution fitting tool, based on the stress distribution histogram, the stress gamma distribution shape parameter and scale parameter of each area to be evaluated can be obtained, as shown in Table 1.

[0101] Table 1

[0102] Partition Partition 1 Partition 2 Partition 3 Partition 4 Shape Parameter a 2.6198 2.4456 1.7779 2.7830 Scale Parameter b 0.5890 0.5480 1.1496 0.6031

[0103] Among them, the input parameter probability distribution model includes the defect morphology parameter distribution model, the fatigue life model parameter distribution model, and the material parameter distribution model. The input parameter probability distribution model is established according to the fatigue life model and the external field defect statistical results, including the following steps:

[0104] S23. Based on the defect uncertainty and the fatigue life model uncertainty, obtain the defect morphology parameters, fatigue life parameters, and material parameters through the external field defect statistical results, and use the normal distribution to respectively fit the distribution mean and variance of the defect morphology parameters, fatigue life parameters, and material parameters to obtain the defect morphology parameter distribution model, the fatigue life model parameter distribution model, and the material parameter distribution model.

[0105] For example, the fatigue life model can adopt the Manson-Coffin formula. The parameters of the fatigue life model include the parameters fitted with the test data and the material-related parameters. The specific fatigue life model is as follows:

[0106]

[0107] Δσ = σ - σ′ f

[0108] Among them, Δε p is the plastic strain amplitude, Δσ is the stress amplitude, E is the elastic modulus, ε′ f is the fatigue ductility coefficient, which represents the initial value of the plastic strain amplitude of the material in a single cycle. It is a parameter obtained through experimental measurement. N f is the life at fatigue failure, that is, the number of cyclic loadings that the material can withstand before failure. c is the fatigue ductility index, usually negative, used to describe the attenuation rate of the strain cycle. σ represents the stress value of the finite element node, and σ′ f fatigue stress index.

[0109] Among them, the defect morphology parameters include the defect depth D and the defect bottom radius R, and the material parameters include the elastic modulus, fatigue ductility coefficient, stress amplitude, and fatigue ductility index.

[0110] Since most statistical data conforms to the Gaussian normal distribution law, this embodiment assumes that the defect morphology parameters, fatigue life parameters, and material parameters satisfy normal independent and identical distribution, where the probability density function of the normal distribution is defined as:

[0111]

[0112] Among them, the calculation formulas for obtaining the mean value and variance are as follows:

[0113]

[0114] In the formula, μ represents the mean parameter of the normal distribution, indicating the central position of the distribution; σ represents the standard deviation parameter of the normal distribution, indicating the degree of dispersion of the distribution, and n represents the total number of samples.

[0115] For example, the defect morphology parameters are obtained from the statistical results of field defects. As shown in Table 2, and then the distribution mean and variance of the defects are obtained by fitting the normal distribution, as Figure 12 shown. In addition, Table 3 shows the data distribution fitting parameters and distribution models of the material parameters. In addition, we obtained the mean defect tolerance of this component from the statistical data as: the depth is 1.75 mm and the radius is 0.6625 mm. It is necessary to evaluate the defect tolerance performance of each area to be evaluated based on this.

[0116] Table 2

[0117] Number 1 2 3 5 6 7 8 Depth D (mm) 2.0 1.2 1.8 1.6 1.5 1.9 2.2 Radius R (mm) 0.8 0.4 0.7 0.7 0.5 0.7 0.9

[0118] Table 3

[0119] Parameter Unit Mean Variance Distribution E MPa 216 0.1 Normal Distribution <![CDATA[ε′ f > 1 -0.2 0.002 Normal Distribution Δσ <![CDATA[×10 -6 Pa]]> 21 2 Normal Distribution C 1 -0.5 0.005 Normal Distribution D mm 1.75 0.3117 Normal Distribution R mm 0.6625 0.1598 Normal Distribution

[0120] S3. Construct a surrogate model sample based on the parameters obtained from the stress probability distribution model and the input parameter probability distribution model of the area to be evaluated, and train and test the fatigue life surrogate model of the area to be evaluated through the surrogate model sample to obtain the trained fatigue life surrogate model.

[0121] Among them, constructing a surrogate model sample based on the parameters obtained from the stress probability distribution model and the input parameter probability distribution model of the area to be evaluated includes the following steps:

[0122] S31. Sample all the stress probability distribution models of the areas to be evaluated to obtain defect location samples, specifically as follows:

[0123] By establishing stress probability distribution models for different regions to be evaluated, Latin hypercube sampling is used to obtain sampling stress values for each region to be evaluated, which form a stress sample. By finding the node number with the minimum mean square error from the extracted stress sample, the spatial position of the defect is obtained. The defect positions of all regions to be evaluated are superimposed to obtain a defect position sample. Among them, the principle of obtaining stress values based on Latin hypercube sampling is as follows:

[0124]

[0125] Among them, σ s is the stress sampling result, and F t -1 is the inverse function of the cumulative distribution function of the stress probability distribution model f t (σ), and U(0,1) represents the data randomly extracted from the uniform distribution of [0,1]. N t is the designed sample number in the t region.

[0126] For example, to obtain the defect position samples of all regions to be evaluated in the tail reduction casing, including: by establishing stress probability distribution models for different regions to be evaluated, the Latin hypercube sampling method, which has better advantages for uniform sampling of small samples, is used to obtain the sampling stress of each region to be evaluated. In addition, based on the minimum mean square error, the stress number with the minimum sampling stress value in each region to be evaluated is found (if there are multiple, the dice-throwing method is used for uniform sampling to obtain the node number), and the nodes of all regions to be evaluated are combined together, which is the defect position sample. As Figure 13 shown, it is the result of defect position sampling obtained based on the above method, where the red dots are the random positions where the defects may appear, used to evaluate the uncertainty of the defects in space. As Figure 13 shown, it is the defect position samples extracted from the four partitions, where the red ones are the nodes drawn, and the black ones are the nodes not drawn.

[0127] S32. Combine and sample the probability quantization model of the input parameters of the surrogate model to obtain the input parameters of the surrogate model, which include defect morphology parameters, fatigue life model parameters, and material parameters.

[0128] S33. Modify the stress of each node in the defect position sample according to the defect morphology parameters and the stress correction factor to obtain the corrected stress of the defect position. Input the maximum corrected stress, fatigue life model parameters, and material parameters into the fatigue life model to obtain the fatigue life sample.

[0129] Since the presence of defects will affect the stress distribution of the structure, and the morphology parameters of the defects will determine this influence degree, as Figure 14As shown in the figure, in the embodiment of the present invention, a stress correction factor is obtained based on the fatigue test data of the small samples with defects. Through the stress correction factor, the stress of each node in the defect position sample is equivalent to the corrected stress including the given defect depth D and the defect bottom radius R, and the corrected stress at the defect position is obtained, specifically as follows:

[0130]

[0131] In the formula, is the corrected stress at the defect position, and λ is the stress correction factor (a weight parameter related to the defect size). In addition, when obtaining the stress correction factor, the depth and radius of the defect need to be converted into standard values, that is, standardization operations are performed respectively:

[0132]

[0133] In the formula, D i represents the original value of the depth of the i-th defect, represents the standardized value of the depth of the i-th defect; R i represents the original value of the radius of the i-th defect, represents the standardized value of the radius of the i-th defect; n represents the total number of defect samples.

[0134] After obtaining the corrected stress at the defect position, the maximum corrected stress, the defect morphology parameters, the fatigue life model distribution parameters, and the material parameters are input into the fatigue life model to obtain the fatigue life samples.

[0135] Since the existence of defects will affect the magnitude of the stress at the defect position nodes, this step proposes a stress correction factor, which comprehensively considers the defect morphology parameters based on the small sample test, as shown in Table 4 specifically. The stress of the defect can be corrected by using the defect morphology parameters obtained by sampling. The stress at the position where the defect is located is scaled through the size of the defect to obtain the corrected stress. Since the corrected stress is not necessarily the maximum stress of the structure, a judgment is still required to output the maximum stress of the structure. Finally, based on the fatigue life model of the defective structure, the fatigue life calculation result of the structure is obtained, and the specific calculation process is as Figure 14 shown.

[0136]

[0137] Table 4

[0138] Using the above steps S31 - S33, the input parameters of the surrogate model are obtained by the combined sampling method for the geometric partitions 1 - 4 of the tail reduction casing. 200 deterministic samples are obtained for each partition based on Latin hypercube sampling respectively. The input parameters of the surrogate model obtained by the combined sampling method are shown in Table 5. The fatigue life sample distribution of each geometric partition calculated by the fatigue life is asFigure 15 , Figure 16 , Figure 17 and Figure 18 as shown.

[0139] Table 5

[0140]

[0141]

[0142] S34. Use the input parameters of the surrogate model as the input and the fatigue life samples as the output to construct the surrogate model samples.

[0143] The surrogate model samples include the input parameters: stress σ at the defect, defect depth, defect radius, Δε p is the plastic strain amplitude, Δσ is the stress amplitude, E is the elastic modulus, ε′ f is the fatigue ductility coefficient, and the output parameter is the fatigue life value in the fatigue life samples.

[0144] Among them, training and testing the fatigue life surrogate model of the area to be evaluated through the surrogate model samples to obtain the trained fatigue life surrogate model includes the following steps:

[0145] S35. Divide the surrogate model samples into a training set and a testing set. The training set is used to train the fatigue life surrogate model, and the testing set is used to verify the fatigue life surrogate model.

[0146] S36. Establish a fatigue life surrogate model for the defective structure according to the genetic algorithm and the Kriging model. Obtain the hyperparameters of the fatigue life surrogate model through the training set, and verify the accuracy of the fatigue life surrogate model through the testing set. If the accuracy meets the requirements, output the fatigue life surrogate model. If the accuracy does not meet the requirements, increase the number of samples in the training set and retrain until the accuracy requirements are met to obtain the trained fatigue life surrogate model, specifically as follows:

[0147] The genetic algorithm adopted is an intelligent optimization method for solving optimization problems, which is used to improve the prediction accuracy of the surrogate model and accelerate the optimization efficiency. Among them, the Kriging model consists of two parts: a polynomial and a Gaussian random process, including:

[0148]

[0149] In the formula, f p (x) represents the p-th polynomial, and its degree is p. β is the hyperparameter to be solved and also the polynomial regression coefficient, which consists of β1, β2, …, β p constitutes, represents the polynomial combination, z(x) represents the Gaussian random process, which follows a normal distribution, that is, z(x) ∼ N(0, σ 2 ), where E[z(x)] represents the mean value of z(x), and D[z(x)] represents the variance of z(x). The correlation function between any two points in the training set is where θ is the correlation coefficient. By introducing the Lagrange multiplier and the least squares method, β can be estimated * =(F T R -1 F) -1 F T R -1 Y.

[0150] For the hyperparameters of the fatigue life surrogate model, it is necessary to clarify the objective function for solving the fatigue life surrogate model. Since z(x) follows a normal distribution, the optimization objective is combined with the natural logarithm function Then, the above formula is differentiated with respect to β * and σ 2 respectively, and then substituted back into the original formula. The problem can be transformed into solving the optimal correlation coefficient θ * , which is defined as:

[0151]

[0152] The optimization process of the genetic algorithm uses a bionic algorithm. That is, using the training set, the input parameters of the surrogate model are input into the fatigue life surrogate model. First, multiple random initial parameters are given during the initial prediction, corresponding to the optimal correlation coefficient θ * , and then the fitness of each individual is calculated using the fitness function. Among them, the fitness function is specifically:

[0153]

[0154] where λ t is the population fitness during the t-th iteration, y pre is the predicted value of the fatigue life surrogate model based on the estimated value of θ t at this time, y true is the true output value, which here corresponds to the fatigue life value of the helicopter transmission system, and n represents that there are n individuals in the population.

[0155] The following optimization process follows the following steps:

[0156] Step1: Randomly generate the initial population, and each population has an estimated value of θ * ;

[0157] Step2: Calculate the fitness of each individual using the fitness function, and form a roulette wheel probability sampling for it. Retain the individuals with high fitness to the next generation, and eliminate the individuals with low fitness. In addition, several gene segments between each individual need to be randomly exchanged, and together with the retained individuals, they form the population of the next generation;

[0158] Step 3: Repeat the process of Step 2 until the fitness function value of the population converges, and select the θ with the highest fitness * as the solution, thereby obtaining the optimal solution of the hyperparameters of the fatigue life surrogate model from the training set.

[0159] Among them, the GA-Kriging fatigue life surrogate model is expressed as: Among them, the input parameters are: x = (Δε p , Δσ, E, ε′ f , σ′ f , c, D, R, σ) T x, Δε p is the plastic strain amplitude, Δσ is the stress amplitude, E is the elastic modulus, ε′ f is the fatigue ductility coefficient, representing the initial value of the plastic strain amplitude in a single cycle, N f is the life at fatigue failure, that is, the number of cyclic loadings that the material can withstand before failure, c is the fatigue ductility exponent, usually negative, used to describe the attenuation rate of the strain cycle.

[0160] Establish the accuracy of the fatigue life surrogate model according to the test set test. If the accuracy requirement is not met, increase the number of training set samples and retrain and verify. The accuracy is defined as the absolute error between the predicted life and the actual life value, and is defined as follows:

[0161] error = |y prediction - y true |

[0162] In the formula, y prediction represents the predicted life value, and y true represents the actual life value.

[0163] For example, divide 200 samples in each area to be evaluated into 160:40 according to 8:2, where 160 are training samples and 40 are test samples. By importing the surrogate model input parameters and fatigue life samples into the established GA-Kriging surrogate model, the fatigue life surrogate model parameters of each area to be evaluated can be obtained. The error convergence curve and the fitness iteration curve of the GA algorithm are as Figure 19 , Figure 20 , Figure 21 and Figure 22 shown. The embodiments of the present invention establish fatigue life surrogate models for different areas to be evaluated for defect tolerance assessment. Table 6 shows the comparison of the fatigue life between the predicted values and the true values in different partitions. It can be seen that the error is small, meeting the requirements of defect tolerance reliability analysis.

[0164] Table 6

[0165]

[0166]

[0167] According to the accuracy of the fatigue life surrogate model established by testing with the test set, if the accuracy requirements are not met, increase the number of samples and retrain and validate. As Figure 23 , Figure 24 , Figure 25 and Figure 26 shown, for the error analysis of the predicted values and actual values of the fatigue life surrogate models of different areas to be evaluated (partition 1-4) of the tail reduction casing, the error band is small, meeting the required modeling accuracy, so reliability assessment can be carried out.

[0168] S4. Based on Bayesian probability fusion, obtain the sampling ratio of each area to be evaluated. According to the sampling ratio, obtain the input parameters of the surrogate model from the probability distribution model of the input parameters of each area to be evaluated, and input the input parameters of the surrogate model into the trained fatigue life surrogate model to obtain the fatigue life distribution of each area to be evaluated, specifically as follows:

[0169] S41. Using finite element software, the area A of the characteristics of each area to be evaluated can be obtained i , and then based on the area, determine the sampling weight of each area to be evaluated, defined as follows:

[0170]

[0171] where r i is the area weight factor of area i to be evaluated, A i is the area of area i to be evaluated, and A is the sum of the surface areas of all areas to be evaluated. Through the above formula, we establish the basic criterion of probability sampling based on area influence factors. Considering the actual outfield statistical results in different areas to be evaluated, the positions and frequencies where defects have occurred. The embodiment of the present invention adopts a probability fusion method based on the Bayesian model, and the Bayesian formula is as follows:

[0172]

[0173] where, A1∪A2∪…∪A n =Ω, P(A|B,C) represents the conditional probability that event A occurs under the condition that events B and C occur simultaneously; P(B|A,C) represents the conditional probability that event B occurs under the condition that events A and C occur simultaneously; P(A,C) represents the joint probability that events A and C occur simultaneously; P(B,C) represents the joint probability that events B and C occur simultaneously; P(A|B) represents the conditional probability that event A occurs under the condition that event B occurs; P(A i ) represents event Ai The prior probability of occurrence; P(B|A i ) represents the conditional probability of event B occurring given that event A i has occurred; P(B) represents the marginal probability of event B occurring; n represents the number of possible values of event A.

[0174] S42. The sampling weights can be biased by using the probability of the appearance of external field statistical defects on the surface, and the sampling quantity of the area to be evaluated is determined based on the biased sampling weights. If the number of defects occurring in different areas to be evaluated is t1, t2, …, t n Then the weight formula based on the external field statistical data is:

[0175]

[0176] In the formula, w i is the statistical weight factor of the area to be evaluated i, t i is the frequency of the appearance of defects in the area to be evaluated i, and T is the sum of the frequencies of the appearance of defects in all areas to be evaluated. And the fusion probability is calculated using the Bayesian probability formula as follows:

[0177]

[0178] The sampling quantity Ni of the area to be evaluated i can be obtained through the above formula i as: Ni i = s i N, where N is the total sampling quantity.

[0179] S43. According to the determined sampling quantity of the area to be evaluated, the input parameters of the surrogate model are obtained from the probability distribution model of the input parameters of each area to be evaluated using Latin hypercube, and substituting them into the trained fatigue life surrogate model, N life distributions can be obtained, and the number of lives in each area to be evaluated is Ni i , so that the life distribution histogram of each area to be evaluated can be obtained by calculating the frequency of each life appearance.

[0180] Taking the tail reduction housing as an example, the surface areas of different areas to be evaluated (partition 1 - 4) of the tail reduction housing are measured by using the finite element surface selection. The measurement results are shown in Table 7. Based on the partition divided by geometric features, the number of defects appearing in each partition is statistically analyzed below based on the external field statistical data, and the statistical results are shown in Table 8. Based on the Bayesian probability fusion formula, the area weight, statistical weight, and sampling weight of the geometric partition are calculated respectively, and the specific results are shown in Table 9.

[0181] Table 7

[0182] Geometric Partition Partition 1 Partition 2 Partition 3 Partition 4 <![CDATA[Surface area (mm 2 )]]> 17353 1992.74 3281.76 7599.1

[0183] Table 8

[0184] Geometric Partition Partition 1 Partition 2 Partition 3 Partition 4 Frequency 4 3 2 1

[0185] Table 9

[0186] Geometric Partition Partition 1 Partition 2 Partition 3 Partition 4 Area Weight 0.5741 0.0659 0.1086 0.2514 Statistical Weight 0.4 0.3 0.2 0.1 Sampling Weight 0.7751 0.0668 0.0733 0.0849

[0187] To obtain the defect tolerance reliability of each area to be evaluated, the average defect tolerance of each area to be evaluated is given above: the defect depth is 1.75 mm and the radius is 0.6625 mm. Then, the embodiments of the present invention use Latin hypercube sampling to obtain 100,000 sets of random surrogate model input parameters, and define the sampling quantities of partitions 1 to 4 according to the sampling weights based on the Bayesian probability fusion formula as: 77510 (partition 1), 6676 (partition 2), 7329 (partition 3), 8486 (partition 4). And import the extracted surrogate model input parameters into the fatigue life surrogate models of each area to be evaluated, obtain the fatigue life distribution of each area to be evaluated, and convert it into a fatigue life probability distribution histogram, as Figure 27 , Figure 28 , Figure 29 and Figure 30 shown.

[0188] S5. Determine the fatigue life value under the given defect tolerance reliability of the area to be evaluated according to the fatigue life distribution of the area to be evaluated; based on the series reliability system theory, determine the defect tolerance reliability of the typical structure according to the fatigue life values under the given defect tolerance reliability of all areas to be evaluated, including the following steps:

[0189] S51. Use the distribution fitting method to fit the established fatigue life probability distribution histograms of each area to be evaluated into a probability density distribution curve (PDF), and convert the probability density distribution curve into a reliability and failure rate curve of fatigue life based on the cumulative integral function, and determine the fatigue life value under the given defect tolerance reliability of the area to be evaluated according to the reliability and failure rate curve of fatigue life.

[0190] As Figure 31 , Figure 32 , Figure 33 and Figure 34 shown, are the fatigue life probability density distribution, the fatigue life reliability and failure rate based on the cumulative integral function for each area to be evaluated. If the indicator function is set to g(x)=N f -2000, it means that as long as the life is greater than 2000 cycles, the tail reduction housing has defect tolerance performance, otherwise it fails to have defect tolerance performance, so as to obtain the reliability of each area to be evaluated, as shown in Table 10.

[0191] S52. Based on the collaborative reliability evaluation theory of the area to be evaluated, determine the defect tolerance reliability of the overall structure of the helicopter transmission system according to the fatigue life value under the given defect tolerance reliability of the area to be evaluated, and complete the defect tolerance reliability analysis of the typical structure of the transmission system.

[0192] In the process of reliability analysis, the life safety is judged through the limit state equation g(x). If g(x)>0, the structure is safe and meets the defect tolerance requirements; otherwise, it does not meet the requirements. By discriminating the life of each area to be evaluated based on the limit state equation, we can obtain:

[0193]

[0194] Among them, p i,f is the failure probability of the area to be evaluated i, and p i,r is the reliability of the area to be evaluated i. G(·) is an indicator function that satisfies g(x)<0. Based on the above formula, the defect tolerance reliability of each area to be evaluated can be obtained. Considering the defect tolerance reliability of the entire structure, based on the following series system reliability theorem:

[0195] R = R1R2…R n = p 1,r p 2,r …p n,r

[0196] In the formula, R1, R2…R n respectively represent the defect tolerance reliabilities of n areas to be evaluated.

[0197] Therefore, the defect tolerance reliability evaluation of the typical structure of the helicopter transmission system based on zoning can be obtained.

[0198] For example, according to the reliability calculation formula of the series system, the overall reliability of the tail reduction gearbox can be obtained as: If the overall reliability is obtained by taking the samples of all areas to be evaluated together, then R = 0.99963. It can be seen that according to the defect tolerance evaluation of zoning collaboration, not only the reliability and failure rate of each area to be evaluated can be obtained, but the overall structure reliability obtained is smaller than that of evaluating all partitions together, which indicates that it is safer in engineering applications.

[0199] Table 10

[0200] Geometric Partition Partition 1 Partition 2 Partition 3 Partition 4 Reliability 0.99995 0.9988 0.9975 0.9994 Failure Rate 0.00005 0.0012 0.0025 0.0006

[0201] Based on the above helicopter transmission system defect tolerance evaluation method, the embodiment of the present invention also provides a helicopter transmission system defect tolerance evaluation system, as Figure 35 shown, including an area division module, a model establishment module, a model training module, a first calculation module, and a second calculation module.

[0202] Among them, the region division module is used to divide the typical structure into multiple regions to be evaluated based on the geometric features of the surface of the typical structure of the helicopter transmission system. The model establishment module is used to establish a stress probability distribution model for each region to be evaluated according to the stress analysis results of the finite element simulation of the defect-free typical structure; establish an input parameter probability distribution model according to the fatigue life model and the statistical results of external field defects.

[0203] The model training module is used to construct a surrogate model sample according to the parameters obtained from the stress probability distribution model and the input parameter probability distribution model of the region to be evaluated, and train and test the fatigue life surrogate model of the region to be evaluated through the surrogate model sample to obtain the trained fatigue life surrogate model.

[0204] The first calculation module is used to obtain the sampling ratio of each region to be evaluated based on Bayesian probability fusion, obtain the surrogate model input parameters from the input parameter probability distribution model of each region to be evaluated according to the sampling ratio, and input the surrogate model input parameters into the trained fatigue life surrogate model to obtain the fatigue life distribution of each region to be evaluated;

[0205] The second calculation module is used to determine the fatigue life value under the given defect tolerance reliability of the region to be evaluated according to the fatigue life distribution of the region to be evaluated; based on the series reliability system theory, determine the defect tolerance reliability of the typical structure according to the fatigue life values under the given defect tolerance reliability of all regions to be evaluated.

[0206] The present invention has been applied to a certain type of helicopter transmission system. The results show that according to the defect tolerance evaluation of partition cooperation, not only can the reliability and failure rate of each partition be obtained, but the overall structure reliability obtained is smaller than that obtained by evaluating all partitions together, which indicates that it is safer during engineering application.

[0207] Although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that: they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements for some of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for evaluating the defect tolerance of a helicopter transmission system, characterized in that, Including: Based on the geometric features of the surface of the typical structure of the helicopter transmission system, the typical structure is divided into multiple regions to be evaluated; According to the stress analysis results of the finite element simulation of the typical structure without defects, a stress probability distribution model for each region to be evaluated is established; an input parameter probability distribution model is established according to the fatigue life model and the statistical results of external field defects; Construct a surrogate model sample according to the parameters obtained from the stress probability distribution model and the input parameter probability distribution model of the region to be evaluated, and train and test the fatigue life surrogate model of the region to be evaluated through the surrogate model sample to obtain the trained fatigue life surrogate model; Based on Bayesian probability fusion, obtain the sampling ratio of each region to be evaluated, obtain the surrogate model input parameters from the input parameter probability distribution model of each region to be evaluated according to the sampling ratio, and input the surrogate model input parameters into the trained fatigue life surrogate model to obtain the fatigue life distribution of each region to be evaluated; According to the fatigue life distribution of the region to be evaluated, determine the fatigue life value under the given defect tolerance reliability of the region to be evaluated; Based on the series reliability system theory, according to the fatigue life values under the given defect tolerance reliability of all regions to be evaluated, determine the defect tolerance reliability of the typical structure.

2. The method for evaluating the defect tolerance of a helicopter transmission system according to claim 1, wherein Based on the geometric features of the surface of the typical structure of the helicopter transmission system, dividing the typical structure into multiple regions to be evaluated includes the following steps: According to the curved surface, splicing and holes on the surface of the typical structure of the transmission system, divide the surface of the typical structure into multiple regions to be evaluated, where each region to be evaluated is the exposed surface outside the typical structure, and there are defects falling in this region to be evaluated in the statistical results of external field defects.

3. The helicopter drive system defect tolerance evaluation method according to claim 1, characterized in that According to the stress analysis results of the finite element simulation of the typical structure without defects, establishing a stress probability distribution model for each region to be evaluated includes the following steps: For the typical structure without defects, through material setting, mesh division and boundary condition loading operations, obtain the finite element simulation results of the typical structure under typical working conditions, and derive the stress results of different regions to be evaluated from the finite element simulation results; Based on the probability and statistics theory, establish a stress distribution histogram for each region to be evaluated, and fit the stress probability distribution model of each region to be evaluated through the gamma distribution.

4. The helicopter drive system defect tolerance assessment method according to claim 1, characterized in that, The input parameter probability distribution model includes a defect morphology parameter distribution model, a fatigue life model parameter distribution model and a material parameter distribution model.

5. The helicopter drive system defect tolerance evaluation method according to claim 4, characterized in that According to the fatigue life model and the statistical results of external field defects, establishing an input parameter probability distribution model includes the following steps: Obtain the defect morphology parameters, fatigue life parameters and material parameters through the statistical results of external field defects, and use the normal distribution to fit the distribution mean and variance of the defect morphology parameters, fatigue life parameters and material parameters respectively to obtain the defect morphology parameter distribution model, the fatigue life model parameter distribution model and the material parameter distribution model.

6. The method for evaluating the defect tolerance of a helicopter transmission system according to claim 5, wherein The defect morphology parameters include defect depth and defect bottom radius, and the material parameters include elastic modulus, fatigue ductility coefficient, stress amplitude and fatigue ductility index.

7. The method for evaluating the defect tolerance of a helicopter transmission system according to claim 1, wherein Constructing a surrogate model sample according to the parameters obtained from the stress probability distribution model and the input parameter probability distribution model of the region to be evaluated includes the following steps: Sample the stress probability distribution model for all regions to be evaluated to obtain defect location samples; Perform combined sampling on the probability quantification model of the input parameters of the surrogate model to obtain the input parameters of the surrogate model, which include defect morphology parameters, fatigue life model parameters, and material parameters; Modify the stress of each node in the defect location sample according to the defect morphology parameters and the stress correction factor to obtain the corrected stress of the defect location. Input the maximum corrected stress, fatigue life model parameters, and material parameters into the fatigue life model to obtain fatigue life samples; Use the input parameters of the surrogate model as inputs and the fatigue life samples as outputs to construct surrogate model samples.

8. The method for evaluating the defect tolerance of a helicopter transmission system according to claim 1, wherein Train and test the fatigue life surrogate model of the region to be evaluated through the surrogate model samples to obtain the trained fatigue life surrogate model, including the following steps: Divide the surrogate model samples into a training set and a test set; Establish a fatigue life surrogate model for the defective structure based on the genetic algorithm and the Kriging model. Obtain the hyperparameters of the fatigue life surrogate model through the training set, and verify the accuracy of the fatigue life surrogate model through the test set. If the accuracy meets the requirements, output the fatigue life surrogate model. If the accuracy does not meet the requirements, increase the number of training set samples and retrain until the accuracy requirements are met to obtain the trained fatigue life surrogate model.

9. The method for evaluating the defect tolerance of a helicopter transmission system according to any one of claims 1-8, characterized in that Based on Bayesian probability fusion, obtain the sampling ratio for each region to be evaluated. According to the sampling ratio, obtain the input parameters of the surrogate model from the probability distribution model of the input parameters for each region to be evaluated. Input the input parameters of the surrogate model into the trained fatigue life surrogate model to obtain the fatigue life distribution for each region to be evaluated, including the following steps: Obtain the area of the characteristics of each region to be evaluated, and then determine the sampling weight for each region to be evaluated based on the area; Bias the sampling weight using the probability of the appearance of defects in the external field statistics on the surface, and determine the sampling quantity for the region to be evaluated based on the biased sampling weight; According to the determined sampling quantity for the region to be evaluated, use Latin hypercube to obtain the input parameters of the surrogate model from the probability distribution model of the input parameters for each region to be evaluated, substitute them into the trained fatigue life surrogate model, obtain multiple life distributions and the quantity of the life for each region to be evaluated, and calculate the frequency of each life appearance to obtain the life distribution histogram for each region to be evaluated.

10. The helicopter drive system defect tolerance assessment method according to claim 9, wherein Determine the fatigue life value under the given defect tolerance reliability for the region to be evaluated according to the fatigue life distribution of the region to be evaluated; Based on the series reliability system theory, determine the defect tolerance reliability of the typical structure according to the fatigue life values under the given defect tolerance reliability for all regions to be evaluated, including the following steps: Use the distribution fitting method to fit the established fatigue life probability distribution histograms of each region to be evaluated into probability density distribution curves, and convert the probability density distribution curves into reliability and failure rate curves of fatigue life based on the cumulative integral function. Determine the fatigue life value under the given defect tolerance reliability for the region to be evaluated according to the reliability and failure rate curves of fatigue life; Based on the collaborative reliability assessment theory of the area to be evaluated, determine the defect tolerance reliability of the overall structure of the helicopter transmission system according to the fatigue life value under the given defect tolerance reliability of the area to be evaluated, and complete the defect tolerance reliability analysis of the typical structure of the transmission system.

11. A helicopter transmission system defect tolerance assessment system, characterized in that, Including: An area division module, which is used to divide the typical structure into multiple areas to be evaluated based on the geometric characteristics of the surface of the typical structure of the helicopter transmission system; A model establishment module, which is used to establish a stress probability distribution model for each area to be evaluated according to the stress analysis results of the finite element simulation of the typical structure without defects; establish an input parameter probability distribution model according to the fatigue life model and the external field defect statistical results; A model training module, which is used to construct a surrogate model sample according to the parameters obtained from the stress probability distribution model and the input parameter probability distribution model of the area to be evaluated, and train and test the fatigue life surrogate model of the area to be evaluated through the surrogate model sample to obtain a trained fatigue life surrogate model; A first calculation module, which is used to obtain the sampling ratio of each area to be evaluated based on Bayesian probability fusion, obtain the surrogate model input parameters from the input parameter probability distribution model of each area to be evaluated according to the sampling ratio, and input the surrogate model input parameters into the trained fatigue life surrogate model to obtain the fatigue life distribution of each area to be evaluated; A second calculation module, which is used to determine the fatigue life value under the given defect tolerance reliability of the area to be evaluated according to the fatigue life distribution of the area to be evaluated; Based on the series reliability system theory, determine the defect tolerance reliability of the typical structure according to the fatigue life values under the given defect tolerance reliability of all areas to be evaluated.

Images