Mechanical component reliability index verification method under virtual-real fusion

Abstract

The application discloses a kind of virtual and real fusion under mechanical component reliability index verification method, belong to reliability engineering technical field.Along with the problem of long test cycle in prior art by physical experiment test mechanical component reliability index, the application is aimed at the highest risk failure mode of mechanical component, and the test bench is built to carry out physical test, the damage value monitoring quantity under the highest risk failure mode is measured, and the preliminary damage model under the highest risk failure mode of mechanical component is established;Then establish damage simulation model, and according to the physical test data and bayesian inference, the unknown parameter random variable in damage simulation model is continuously updated, and the parameterized damage simulation model is obtained;Finally, according to the reliability index under the highest risk failure mode of mechanical component, the parameterized damage simulation model is compared and verified.The method can effectively reduce the sample required when mechanical component reliability index verification is carried out, reduce test cost and test cycle.

Description

Technical Field

[0001] The present invention relates to the field of reliability engineering technology, and in particular to a method for verifying reliability indicators of mechanical parts under virtual-real integration. Background Art

[0002] Mechanical components, particularly those used in aerospace equipment, have extremely high reliability requirements. Verifying these reliability requirements often requires a large number of samples, leading to high testing costs. For mechanical components with a predominantly wear-out failure mode, this also poses the challenge of long testing cycles.

[0003] Taking aircraft landing gear as an example, as the number of retraction and extension cycles increases, the hinges at the rod connections wear deeper, affecting the normal retraction and extension of the landing gear, leading to failure modes such as jamming and precision failure, seriously affecting flight safety. Furthermore, measuring the wear of the hinges connecting the landing gear retraction mechanism is difficult, making direct damage monitoring impossible during takeoff and landing, resulting in a lag in the verification of its reliability indicators.

[0004] Existing methods for verifying the reliability of mechanical components mostly rely on physical experiments, such as fatigue testing and damage testing. However, for long-life or high-intensity testing, these tests can be time-consuming and resource-intensive. Therefore, a virtual-physical hybrid method for verifying the reliability of mechanical components is urgently needed to address these issues. Summary of the Invention

[0005] In response to the above-mentioned problems, the present invention aims to provide a method, system and equipment for verifying the reliability indicators of mechanical parts under the fusion of virtual and real, so as to solve the difficult problems of large sample requirements, high test costs and long test cycles when verifying the reliability indicators of mechanical parts.

[0006] In order to achieve the above object, the technical solution adopted by the present invention is as follows:

[0007] On the one hand, the present invention provides a method for verifying the reliability index of mechanical parts under virtual-real fusion, comprising the following steps:

[0008] S1: Establish a preliminary damage model for the highest risk failure mode of mechanical components;

[0009] S2: constructing a damage simulation model based on the preliminary damage model in step S1, and characterizing the unknown parameter random variables in the damage simulation model through physical experimental data and Bayesian inference to obtain a parameterized damage simulation model;

[0010] S3: Verify the parameterized damage simulation model obtained in step S2 based on the reliability index of the mechanical component under the highest risk failure mode.

[0011] Furthermore, the specific operations of step S1 include the following steps:

[0012] S101: Identify the highest risk failure modes of mechanical components and their corresponding influencing factors;

[0013] S102: Determine the reliability index under the highest risk failure mode, and measure the damage value monitoring quantity under the highest risk failure mode through a physical test platform;

[0014] S103: Determine the uncertainty information of the unknown parameter random variable in the influencing factors in step S101;

[0015] S104: Establish a preliminary damage model for the highest risk failure mode of mechanical components.

[0016] Furthermore, the unknown parameter random variables in step S103 include unknown parameter random variables whose distribution information can be determined, unknown parameter random variables with prior information, and unknown parameter random variables without any prior information.

[0017] Furthermore, the specific operation of step S2 includes the following steps:

[0018] S201: Using the preliminary damage model constructed in step S1 as the initial state of the damage simulation model;

[0019] S202: Simulating unknown parameter random variables in the damage simulation model through physical experimental data and Bayesian inference, thereby updating the damage simulation model;

[0020] S203: After determining the representation method of the unknown parameter random variable, a parameterized damage simulation model is finally obtained.

[0021] Furthermore, the specific operation of step S202 includes the following steps:

[0022] S2021: Random variables with unknown parameters for the highest risk failure mode K Perform Bayesian update and use the K The posterior distribution of the unknown parameter random variable is used as the prior distribution of the current period K Bayesian update of :

[0023] ;

[0024] in, h Indicates the damage value monitoring quantity under the current damage cycle under the highest risk failure mode, is the unknown parameter random variable in the current damage cycle K The prior distribution of Indicates that when there is new physical test platform to transmit damage value monitoring data, the parameter Perform Bayesian update of the posterior distribution and use it as the prior distribution for the next parameter update; P ( h ) is the marginal distribution of the damage value monitoring data after processing, which is used for normalization so that the integral of the posterior distribution is 1. P ( h|K ) is the likelihood function of the damage value monitoring data after processing;

[0025] S2022: n In each damage cycle, the damage value monitoring quantity of each damage cycle is defined as h i , i =1,2,…, n , it is stipulated that the difference between the actual damage value monitored in the current damage cycle and the damage value predicted by the prior distribution has a mean of 0 and a standard deviation of Normal distribution, then:

[0026] ;

[0027] in, To use the unknown parameter random variable of the current period K The damage amount prediction value of the current cycle based on the prior distribution of Indicates the i Likelihood function of the damage value monitoring data after processing within a damage cycle;

[0028] S2023: When the relative error between the damage value prediction value and the physical test damage value monitoring value of two consecutive damage cycles is less than 0.1%, the Bayesian update of the unknown parameter random variable is stopped, and the posterior distribution of the unknown parameter random variable obtained in the last cycle is used as the final result;

[0029] S2024: Repeat steps S2021-S2023, and each group completes the Bayesian update of the unknown parameter random variable K The mean and standard deviation are u i and σ i , i =1,2,…, n , then the unknown parameter random variable K The mean is , the standard deviation is Normal distribution.

[0030] Furthermore, the specific operation of step S3 includes the following steps:

[0031] S301: Establishing a function for reliability analysis of mechanical components based on the reliability index under the highest risk failure mode;

[0032] S302: Using the probability density function of the parameters of the influencing factors under the highest risk failure mode of mechanical components , using the Latin hypercube method to generate Monte Carlo pooling of samples ;

[0033] S303: Using the Kriging proxy model to update the data in the sample pool to obtain the final Kriging proxy model;

[0034] S304: Predict and compare the reliability indicators of mechanical parts based on the final Kriging proxy model.

[0035] Furthermore, the function of the reliability analysis of mechanical parts in step S301 is:

[0036] ;

[0037] Where, is the critical value of damage failure, The actual amount of damage to a component when it reaches its design life; It is considered safe, It is deemed to be invalid.

[0038] Furthermore, the specific operation of step S303 includes the following steps:

[0039] S3031: In the Monte Carlo sample pool Select the initial training sample set Train the Kriging proxy model to obtain the initial Kriging proxy model;

[0040] S3032: The initial Kriging proxy model outputs a prediction value with confidence based on the input sample point data. The sample points whose difference between the absolute value of the predicted mean and the failure surface is less than the threshold are added to the sample training set to update the Kriging proxy model until the learning function value of all prediction values ​​in the specified sample range is , stop updating the Kriging model, and use the last updated Kriging proxy model as the final Kriging proxy model.

[0041] Furthermore, the learning function described in step S3032 is:

[0042] ;

[0043] Where, The function predicts the mean value of the input sample point. is the standard deviation of the predicted values.

[0044] On the other hand, the present invention also provides a reliability index verification system for mechanical parts under virtual-real fusion, including a physical test module, a damage simulation model module and a reliability index verification module;

[0045] The physical test module is used to collect physical test data and establish a preliminary damage model under the highest risk failure mode of mechanical components;

[0046] The damage simulation model module is used to build a parameterized damage simulation model;

[0047] The reliability index verification module is used to verify the parameterized damage simulation model;

[0048] Among them, the physical test module, damage simulation model module and reliability index verification module are all implemented using the reliability index verification method of mechanical parts under the virtual-reality fusion as described above.

[0049] The beneficial effects of the present invention are:

[0050] The present invention provides a method for verifying the reliability index of mechanical parts under the fusion of virtual and real. The method first sets up a test bench for the highest risk failure mode of the mechanical parts to conduct physical tests, measures the damage value monitoring quantity under the highest risk failure mode, and establishes a preliminary damage model under the highest risk failure mode of the mechanical parts; then establishes a damage simulation model, and continuously updates the unknown parameter random variables in the damage simulation model based on the physical test data and Bayesian inference to obtain a parameterized damage simulation model; finally, the parameterized damage simulation model is compared and verified based on the reliability index under the highest risk failure mode of the mechanical parts. The method of the present invention can effectively reduce the samples required for reliability index verification of mechanical parts, reducing the test cost and test cycle. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 This is a flow chart of the method for verifying reliability indicators of mechanical parts under virtual-reality fusion in the present invention.

[0052] Figure 2 This is the experimental process of the simulation experiment in the present invention.

[0053] Figure 3 This is a finite element model diagram of the hinge in the simulation test of the present invention.

[0054] Figure 4 Schematic diagram of the relative motion relationship between the sleeve and the shaft in the simulation test of the present invention.

[0055] Figure 5 This is a simulation result diagram of the simulation test of the present invention.

[0056] Figure 6 This is the overall framework diagram of the mechanical parts reliability index verification system under the virtual-reality fusion in the present invention. DETAILED DESCRIPTION

[0057] In order to enable those skilled in the art to better understand the technical solution of the present invention, the technical solution of the present invention is further described below in conjunction with the accompanying drawings and embodiments.

[0058] Example 1

[0059] Embodiment 1 provides a method for verifying reliability indicators of mechanical parts under virtual-real fusion, which specifically includes the following steps:

[0060] S1: Establish a preliminary damage model for the highest risk failure mode of mechanical components;

[0061] Specifically, S101: determine the highest risk failure mode of mechanical components and the corresponding influencing factors;

[0062] The functional requirements, structural composition, and boundary conditions of the mechanical component to be analyzed are analyzed, and the highest failure modes and corresponding influencing factors of the mechanical component are determined using Failure Mode and Effects Analysis (FMEA) and Fault Tree Analysis (FTA) techniques. More specifically, FMEA is used to systematically identify and list the potential failure modes of each component of the mechanical component. The severity (S), probability of occurrence (O), and difficulty of detection (D) of each potential failure mode are evaluated, and the Risk Priority Number (RPN) (RPN = S × O × D) is calculated to determine the failure mode that requires priority attention, i.e., the highest-risk failure mode. For the highest-risk failure mode identified in the FMEA, the Fault Tree Analysis (FTA) technique is further applied to construct a fault tree at the system level to identify the influencing factors corresponding to the highest-risk failure mode.

[0063] S102: Determine the reliability index under the highest risk failure mode, and measure the damage value monitoring quantity under the highest risk failure mode through a physical test platform;

[0064] Based on the highest-risk failure mode determined in step S101, determine the reliability index under the highest-risk failure mode. Build a physical test platform, and select damage value monitoring quantities based on the influencing factors of the highest-risk failure mode. For example, the damage value monitoring quantity for the wear failure mode can be selected as the wear depth or wear volume within the operating cycle, and the damage value monitoring quantity for the fatigue failure mode can be selected as the number of cycles without cracks within a specified time. After selecting the damage value monitoring quantity, measure the damage value monitoring quantity under the highest-risk failure mode using the physical test platform. The damage value monitoring quantity is used to verify the reliability index.

[0065] S103: Determine the uncertainty information of the unknown parameter random variable in the influencing factors in step S101;

[0066] According to the highest risk failure mode and influencing factors determined in step S101, it is further necessary to determine the uncertainty information of the unknown parameter random variables in the influencing factors; the unknown parameter random variables are divided into three types: (1) unknown parameter random variables whose distribution information can be determined based on manuals and engineering experience, such as the geometric dimensions of the mechanism and the elastic modulus of commonly used materials; (2) unknown parameter random variables with prior information, such as the wear coefficient of the material; (3) unknown parameter random variables without any prior information.

[0067] The first type of unknown parameter random variables can be determined through inquiry or engineering experience, such as the Young's modulus, Poisson's ratio, manufacturing error, etc. of the material; the second type of unknown parameter random variables have prior information. Their prior distribution is first determined based on previous test data, and then limited physical test data is used to perform statistical decision-making and Bayesian estimation. The posterior distribution is continuously updated based on the likelihood function of the prior distribution and test data, such as the wear coefficient of the material; the third type of parameters do not have any prior information. The non-informative prior method is used to perform Bayesian inference using limited physical test data to obtain the posterior distribution of this type of parameters.

[0068] S104: Establish a preliminary damage model for the highest risk failure mode of mechanical components;

[0069] Based on the initial information of the influencing factors determined in step S101, a preliminary damage model under the highest risk failure mode is established.

[0070] For example, the initial damage model of the wear failure mode adopts the Archard wear model, which is expressed as:

[0071] ;

[0072] Where, V is the wear volume, S is the sliding distance of the wear area,k is the material wear coefficient, F is the normal load size, H is the material hardness. The commonly used wear failure mode monitoring quantity is selected as the wear depth instead of the wear volume, so the Archard wear model can be rewritten as:

[0073] ;

[0074] The Archard wear model is further deformed as:

[0075] ;

[0076] wherein, A is the wear surface area, h is the wear depth, is the contact stress of the contact point of the two objects, is the linear wear coefficient.

[0077] It should be noted here that the preliminary damage model under different risk highest failure modes is different, and the preliminary damage model under each risk highest failure mode is specific, which belongs to the prior art. The above example illustrates the preliminary damage model of the wear failure mode, and the preliminary damage model under other risk highest failure modes is not described in detail in the present application.

[0078] Further, step S2: constructing a damage simulation model based on the preliminary damage model in step S1, and characterizing the unknown parameter random variable in the damage simulation model to obtain a parameterized damage simulation model.

[0079] Specifically, S201: taking the preliminary damage model constructed in step S1 as the initial state of the damage simulation model;

[0080] S202: simulating the unknown parameter random variable in the damage simulation model, thereby updating the damage simulation model; setting the transmission time interval as T, transmitting the damage value monitoring quantity of the mechanical component every time T to the damage simulation model, using the posterior distribution of the unknown parameter random variable in the last period as the prior distribution of the current period, and jointly performing Bayesian updating with the damage value monitoring quantity data of the physical test platform in the current period to obtain the posterior distribution of the unknown parameter random variable.

[0081] More specifically, the process of Bayesian updating includes the following steps: Bayesian updating is obtained by combining the prior distribution of the unknown parameter random variable and the likelihood function of the newly obtained damage value monitoring quantity data to obtain a joint probability distribution, and then normalizing the joint probability distribution to obtain the posterior distribution of the unknown parameter random variable, wherein the Bayesian theorem is the core of the Bayesian updating method. The Bayesian theorem is specifically represented as:

[0082] ;

[0083] Where, is the posterior distribution of the random variable with unknown parameters, is the likelihood function after damage value monitoring data processing, is the prior distribution of the random variable with unknown parameters. is the marginal distribution of the damage value monitoring data after processing, which is used for normalization so that the integral of the posterior distribution is 1.

[0084] For the unknown parameter random variable under the highest risk failure mode K Perform Bayesian updating, K With a prior distribution (if K Without any prior information, a non-information prior method is used, that is, a distribution is assigned to the parameter as a prior distribution), using the previous cycle obtained K The posterior distribution of the unknown parameter random variable is used as the prior distribution of the current period K The Bayesian update of , then:

[0085] ;

[0086] in, h Indicates the damage value monitoring quantity under the current damage cycle under the highest risk failure mode, is the unknown parameter random variable in the current damage cycle K The prior distribution of Indicates that when there is new physical test platform to transmit damage value monitoring data, the parameter Perform Bayesian update of the posterior distribution and use it as the prior distribution for the next parameter update; P ( h ) is the marginal distribution of the damage value monitoring data after processing, which is used for normalization so that the integral of the posterior distribution is 1. P ( h|K ) is the likelihood function after damage value monitoring data processing.

[0087] existing n damage cycles, and the damage value monitoring quantity of each damage cycle is , it is stipulated that the difference between the actual damage value monitored in the current damage cycle and the damage value predicted by the prior distribution has a mean of 0 and a standard deviation of Normal distribution, then:

[0088] ;

[0089] in, To use the unknown parameter random variable of the current period K Prior distribution of the key parameters of the previous period K The posterior distribution of the damage amount of the current cycle is predicted (mean prediction), and K It is a linear relationship, and the size of the linear relationship is determined by the relevant characterization parameters. Correspondingly, Indicates the i Likelihood function of the damage value monitoring data after processing within a damage cycle;

[0090] When the relative error between the predicted value of the damage amount (calculated using the mean value) of two consecutive damage cycles and the monitored value of the physical test damage value is less than 0.1%, the Bayesian update of the unknown parameter random variable is stopped, and the posterior distribution of the unknown parameter random variable obtained in the last cycle is used as the final result.

[0091] Repeat the above steps n Group experiment, each group completes the Bayesian update after the unknown parameter random variable K The mean and standard deviation are u i and σ i ( i =1,2,…, n ), then the unknown parameter random variable K The mean is , the standard deviation is Normal distribution.

[0092] S203: After determining the representation method of the unknown parameter random variable, a parameterized damage simulation model is finally obtained.

[0093] Furthermore, step S3: verifying the damage simulation model in step S2 according to the reliability index of the mechanical component under the highest risk failure mode.

[0094] Specifically, S301: Based on the reliability index under the highest risk failure mode, a function for reliability analysis of mechanical parts is established:

[0095] ;

[0096] Where, is the critical value (threshold) of damage failure, is the actual damage amount of the component when it reaches its design life. It is considered safe, It is considered as failure. For example, the critical value of wear failure mode is the wear depth or wear volume, the actual damage amount The actual wear depth or wear volume of the mechanical part when reaching the designed service life. In the present application, the actual damage amount d According to the damage simulation model in step S2.

[0097] S302: Generate a Monte Carlo sample pool: use the probability density function of the parameters in the finally determined influencing factors to generate a Monte Carlo sample pool containing samples using the Latin hypercube method . The number of samples in the sample pool is determined according to the reliability index requirement. Generally, the total number of samples , is the failure probability specified for the mechanical part.

[0098] S303: Update the data in the sample pool using the Kriging surrogate model;

[0099] First, select an initial training sample set from the Monte Carlo sample pool to train the Kriging surrogate model. The relevant model uses a Gaussian process model, and the regression model uses a constant to obtain the initial Kriging surrogate model.

[0100] Then, update the initial Kriging surrogate model. The Kriging surrogate model is based on the fact that the functional function value solved by the input sample obeys a normal distribution, and the mean value is defined as the distance from the failure surface, and the standard deviation is the uncertainty of the predicted value. That is, the initial Kriging surrogate model outputs a predicted value with a confidence level based on the input sample point data. In the present application, the learning function is defined as:

[0101] ;

[0102] In the formula, is the predicted mean value of the input sample point function function, is the standard deviation of the predicted value. When the predicted mean value is the same, the larger the standard deviation, the larger the learning function value; when the predicted standard deviation is the same, the smaller the absolute value of the mean, the larger the learning function value; the sample points with a predicted mean absolute value less than a threshold value (the specific selection of the threshold value can be set in advance according to different mechanical parts, such as 0.05) and a large standard deviation (greater than 0.01) are added to the sample training set to update the Kriging surrogate model. Generally, the is the judgment basis for stopping updating the Kriging surrogate model. If the learning function value of the sample point in the prediction process , then the sample point is added to the training set Re-update the Kriging surrogate model until the learning function value of all predicted values ​​in the specified sample range is , stop updating the Kriging model, and use the last updated Kriging proxy model as the final Kriging proxy model.

[0103] S304: Based on the final Kriging proxy model obtained in step S303, all sample points are substituted into the final Kriging proxy model to calculate the performance function value, and the Monte Carlo digital simulation method is used to solve the reliability prediction value of the mechanical component under the specified failure mode. (Monte Carlo digital simulation method to solve the reliability prediction value This is prior art and will not be described in detail in this invention. In this invention, the Monte Carlo digital simulation method generates a large number of samples for calculating reliability based on the law of large numbers. The number of samples in the overall sample pool to be generated is based on the number of samples in the sample pool. The determination is determined by the determination method, see step S302.

[0104] Use the final Kriging surrogate model to predict all samples in the population and output the reliability prediction value of the target mechanical parts under the highest risk failure mode , failure probability prediction value and its coefficient of variation ,like , then the reliability of the prediction is considered to be Valid; if , then the number of samples in the overall sample pool needs to be expanded N , repeat step S303 until , and finally output the result.

[0105] The reliability prediction value , failure probability prediction value Compare and verify with the reliability index of mechanical parts. If the reliability prediction value Less than the reliability index and failure probability prediction value required by mechanical parts If the reliability index is greater than the required reliability index of the mechanical component, then the design of the mechanical component is unqualified; if the reliability prediction value Greater than the reliability index and failure probability prediction value required by mechanical components If the reliability index is less than that required by the mechanical parts, then the mechanical parts are considered to be qualified in design.

[0106] Simulation test:

[0107] This simulation test uses pin and bushing wear (hinge wear) as an example. A geometric model is established based on the actual dimensions of the hinge. A finite element model is then built. Based on the actual working conditions of the hinge, a load is applied to the hinge and the speed of the pin relative to the bushing is set to predict the wear depth of the hinge and verify its reliability.

[0108] The specific steps of the simulation test are as follows: Figure 2 As shown, the following steps are included:

[0109] (1) Establishing a finite element simulation model

[0110] The finite element simulation model needs to set physical geometric parameters and boundary conditions. The finite element model is established in the simulation software according to the actual physical model parameters of the pin and sleeve, as shown in the attached figure. Figure 3 shown.

[0111] The geometric dimensions and material selection of the pin and bushing: The finite element model unit type is selected as a two-dimensional plane, the radius of the pin is set to 5.99mm, the inner radius of the bushing is set to 6mm, and the outer radius is set to 10mm, as shown in the attached figure. Figure 4 As shown in the figure, the material of the pin is QSn8-0.3, and the material of the sleeve is 30CrMnSiA. After the material is selected, the properties of the material are automatically obtained in the finite element simulation software.

[0112] Speed ​​and direction of relative motion between the pin and bushing: Set the pin's rotation direction to counterclockwise, the angular velocity to 18.326 rad / s, and fix the position of the bushing to keep it stationary.

[0113] Apply boundary conditions and loads: Select the contact area between the sleeve and the pin, define the friction coefficient of the contact area, set it to 0.12, apply the load at the geometric center of the pin, and always in the vertical upward direction. The applied load is 100N.

[0114] (2) Add wear subroutine

[0115] The wear simulation model uses a wear subroutine written in Fortran. The subroutine contains the Archard wear model and the wear coefficient that needs to be updated by Bayesian. The initial value of the wear coefficient is selected as the priori initial value, which is 9. , associate the subroutine to the simulation model.

[0116] (3) Set simulation duration and grid division

[0117] After completing the above steps, refine the mesh in the wear area of ​​the pin and bushing, and select the wear nodes in the outermost layer of the wear area. Set the total simulation time according to the requirements , set the step size to , and conduct simulation experiments.

[0118] (4) Update the wear coefficient

[0119] In each During the step time, the wear area calculates the wear depth based on the load and slip distance at the node, and combines it with the actual wear depth transmitted from the physical test module to perform a Bayesian update of the wear coefficient. Each time it is updated, the mesh of the finite element simulation model is reconstructed and the above steps are repeated. If the Bayesian update of the wear coefficient does not meet the stopping criteria, increase the simulation test time. Repeat the above steps until the wear coefficient meets the criterion for stopping updating and output the final wear coefficient.

[0120] (5) Determine the simulation model

[0121] After the wear coefficient is updated and determined through the above steps, the simulation model is used instead of the physical test module to predict the wear depth and set the total simulation time. , step length , a total of ten wear cycles, the final maximum wear depth result is obtained, as shown in the attached Figure 5 shown.

[0122] From the attached Figure 5 It can be seen from the simulation results that the damage simulation model in the present invention can accurately reconstruct the wear morphology of the contact interface and effectively predict the wear amount in the contact area, providing a theoretical basis for the reliability assessment and maintenance decision-making of the mechanical system.

[0123] Example 2

[0124] The second embodiment provides a mechanical parts reliability index verification system under virtual-real fusion, as shown in the attached Figure 6 As shown, it includes physical test module, damage simulation model module and reliability index verification module;

[0125] The physical test module is used to collect physical test data and establish a preliminary damage model under the highest risk failure mode of mechanical components;

[0126] The damage simulation model module is used to build a parameterized damage simulation model;

[0127] The reliability index verification module is used to verify the parameterized damage simulation model;

[0128] Among them, the physical test module, the damage simulation model module and the reliability index verification module are all implemented using the reliability index verification method of mechanical parts under the virtual-real fusion described in Example 1.

[0129] The basic principles, main features, and advantages of the present invention are shown and described above. Those skilled in the art should understand that the present invention is not limited to the foregoing embodiments. The foregoing embodiments and descriptions are merely illustrative of the principles of the present invention. Various changes and modifications may be made to the present invention without departing from the spirit and scope of the present invention. Such changes and modifications are intended to fall within the scope of the present invention. The scope of protection claimed in the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for verifying the reliability index of mechanical parts under virtual-real integration, characterized by: The following steps are involved: S1: Establish a preliminary damage model for the highest risk failure mode of mechanical components; S2: constructing a damage simulation model based on the preliminary damage model in step S1, and characterizing the unknown parameter random variables in the damage simulation model through physical experimental data and Bayesian inference to obtain a parameterized damage simulation model; S3: Validate the parameterized damage simulation model obtained in step S2 based on the reliability index of the mechanical component under the highest risk failure mode; The specific operations of step S2 include the following steps: S201: Using the preliminary damage model constructed in step S1 as the initial state of the damage simulation model; S202: Simulating unknown parameter random variables in the damage simulation model through physical experimental data and Bayesian inference, thereby updating the damage simulation model; S203: After determining the representation method of the unknown parameter random variable, a parameterized damage simulation model is finally obtained; The specific operation of step S202 includes the following steps: S2021: Random variables with unknown parameters for the highest risk failure mode K Perform Bayesian update and use the K The posterior distribution of the unknown parameter random variable is used as the prior distribution of the current period K Bayesian update of : ； in, h Indicates the damage value monitoring quantity under the current damage cycle under the highest risk failure mode, is the unknown parameter random variable in the current damage cycle K The prior distribution of Indicates that when there is new physical test platform to transmit damage value monitoring data, the parameter Perform Bayesian update of the posterior distribution and use it as the prior distribution for the next parameter update; P ( h ) is the marginal distribution of the damage value monitoring data after processing, which is used for normalization so that the integral of the posterior distribution is 1. P ( h|K ) is the likelihood function of the damage value monitoring data after processing; S2022: n In each damage cycle, the damage value monitoring quantity of each damage cycle is defined as h i , i =1,2,…, n , it is stipulated that the difference between the actual damage value monitored in the current damage cycle and the damage value predicted by the prior distribution has a mean of 0 and a standard deviation of Normal distribution, then: ； in, To use the unknown parameter random variable of the current period K The damage amount prediction value of the current cycle based on the prior distribution of Indicates the i Likelihood function of the damage value monitoring data after processing within a damage cycle; S2023: When the relative error between the damage value prediction value and the physical test damage value monitoring value of two consecutive damage cycles is less than 0.1%, the Bayesian update of the unknown parameter random variable is stopped, and the posterior distribution of the unknown parameter random variable obtained in the last cycle is used as the final result; S2024: Repeat steps S2021-S2023, and each group completes the Bayesian update of the unknown parameter random variable K The mean and standard deviation are u i and σ i , i =1,2,…, n , then the unknown parameter random variable K The mean is , the standard deviation is Normal distribution; The specific operation of step S3 includes the following steps: S301: Establishing a function for reliability analysis of mechanical components based on the reliability index under the highest risk failure mode; S302: Using the probability density function of the parameters of the influencing factors under the highest risk failure mode of mechanical components , using the Latin hypercube method to generate Monte Carlo pooling of samples ; S303: Using the Kriging proxy model to update the data in the sample pool to obtain the final Kriging proxy model; S304: Predict and compare the reliability indicators of mechanical parts based on the final Kriging proxy model.

2. The method for verifying reliability index of mechanical parts under virtual-reality fusion according to claim 1 is characterized in that: The specific operation of step S1 includes the following steps: S101: Identify the highest risk failure modes of mechanical components and their corresponding influencing factors; S102: Determine the reliability index under the highest risk failure mode, and measure the damage value monitoring quantity under the highest risk failure mode through a physical test platform; S103: Determine the uncertainty information of the unknown parameter random variable in the influencing factors in step S101; S104: Establish a preliminary damage model for the highest risk failure mode of mechanical components.

3. The reliability index verification method of mechanical parts under virtual-reality fusion according to claim 2 is characterized in that: The unknown parameter random variables in step S103 include unknown parameter random variables whose distribution information can be determined, unknown parameter random variables with prior information, and unknown parameter random variables without any prior information.

4. The reliability index verification method of mechanical parts under virtual-reality fusion according to claim 3 is characterized in that: The function of the reliability analysis of mechanical parts in step S301 is: ； Where, is the critical value of damage failure, The actual amount of damage to a component when it reaches its design life; It is considered safe, It is deemed to be invalid.

5. The method for verifying reliability index of mechanical parts under virtual-reality fusion according to claim 4 is characterized in that: The specific operation of step S303 includes the following steps: S3031: In the Monte Carlo sample pool Select the initial training sample set Train the Kriging proxy model to obtain the initial Kriging proxy model; S3032: The initial Kriging proxy model outputs a prediction value with confidence based on the input sample point data. The sample points whose difference between the absolute value of the predicted mean and the failure surface is less than the threshold are added to the sample training set to update the Kriging proxy model until the learning function value of all prediction values ​​in the specified sample range is , stop updating the Kriging model, and use the last updated Kriging proxy model as the final Kriging proxy model.

6. The reliability index verification method of mechanical parts under virtual-reality fusion according to claim 5 is characterized in that: The learning function described in step S3032 is: ； Where, The function predicts the mean value of the input sample point. is the standard deviation of the predicted values.

7. The reliability index verification system of mechanical parts under the integration of virtual and real is characterized by: Including physical test module, damage simulation model module and reliability index verification module; The physical test module is used to collect physical test data and establish a preliminary damage model under the highest risk failure mode of mechanical components; The damage simulation model module is used to build a parameterized damage simulation model; The reliability index verification module is used to verify the parameterized damage simulation model; Among them, the physical test module, the damage simulation model module and the reliability index verification module are all implemented by the reliability index verification method of mechanical parts under virtual-reality fusion as described in any one of claims 1-6.

Images