A method and system for rapid prediction of high-cycle fatigue life and dangerous sites of a metal component

Abstract

This invention discloses a rapid prediction method and system for high-cycle fatigue life and critical locations of metal components, belonging to the field of structural health monitoring technology. It aims to address the shortcomings of existing high-cycle fatigue life assessment methods, which rely heavily on finite element simulations and critical point selection, resulting in low computational efficiency and difficulty in obtaining the full-field life distribution. The method includes: acquiring input information of the component to be predicted; inputting the input information into a pre-constructed nonlinear mapping model to obtain the component's life field and the location coordinates of the shortest-life element. By pre-constructing a nonlinear mapping model offline, this invention allows for rapid output of the full-field life field and the shortest-life location by only requiring structural and operating condition information during the prediction stage. This significantly reduces computation time and resource consumption, making it suitable for rapid comparison and optimization design of multiple operating conditions and structural schemes.

Description

Technical Field

[0001] This invention relates to a method and system for rapid prediction of high-cycle fatigue life and critical locations of metal components, belonging to the field of structural health monitoring technology. Background Technology

[0002] As the service life of metal components increases, due to various unreasonable factors in the design, manufacturing and operation management of metal components, the long-term cyclic action of service loads, and the deterioration of the high-cycle fatigue performance of the materials themselves, metal components will inevitably expose various inherent or acquired fatigue defects and hidden dangers, posing a threat to the safe operation of the structure.

[0003] Existing engineering projects typically use methods such as strain testing, acoustic emission, ultrasonic testing, and magnetic particle testing to monitor the fatigue state and detect defects in metal components. The signals obtained by these methods carry information to some extent about the structural stress state, damage evolution, and changes in material properties. However, due to limitations in the number of sensors, testing conditions, and signal processing technology, the interpretation of the observed signals remains difficult. The test results can mostly only provide qualitative or semi-quantitative judgments on the fatigue status of the overall structure or local areas. It is difficult to accurately obtain the fatigue life distribution of various parts of the structure under complex working conditions and the precise coordinates of critical parts. It is also not conducive to the rapid comparison and evaluation of various structural schemes and load conditions during the design phase.

[0004] While existing numerical calculation methods based on computational mechanics can establish finite element models, solve stress or stress amplitude fields under given boundary conditions and loads, and then analyze the fatigue life of structures by combining the SN curve theory of high-cycle fatigue, they often have limitations in practical applications, such as high computational cost, reliance on manual experience for mesh generation and boundary condition processing, and difficulty in efficiently carrying out large-scale life calculations under various combinations of structural parameters and working conditions. In addition, they can usually only perform life assessments for a few pre-selected critical sections or monitoring points, and it is difficult to quickly obtain the life field of all elements of the structure and the spatial location of the shortest-lived element. Summary of the Invention

[0005] The technical problem to be solved by the present invention is to provide a rapid prediction method and system for high-cycle fatigue life and critical parts of metal components, overcoming the shortcomings of existing high-cycle fatigue life assessments, such as reliance on a large number of finite element simulations and selection of critical points, low computational efficiency, and difficulty in obtaining the full-field life distribution.

[0006] To achieve the above objectives, the present invention employs the following technical solution:

[0007] In a first aspect, the present invention provides a method for rapid prediction of high-cycle fatigue life and critical locations of metal components, comprising:

[0008] The input information of the metal component to be predicted is obtained and input into a pre-constructed nonlinear mapping model to obtain the life field and the position coordinates of the shortest life element of the metal component. The input information includes the geometric dimensions of the metal component, boundary conditions, load application location, load type, load magnitude and material parameters. The life field and the position coordinates of the shortest life element of the metal component are respectively the high-cycle fatigue life distribution and the critical location of the metal component.

[0009] The construction process of the nonlinear mapping relationship model includes:

[0010] Several pairs of datasets are randomly generated, wherein the generation of any pair of datasets includes:

[0011] Obtain the information parameters of the target metal component and establish a geometric model of the component; the information parameters include the geometric dimension parameters, boundary condition parameters, load application location parameters, load form parameters, load magnitude parameters, and material parameters of the metal component;

[0012] The component's geometric model is divided into mesh elements using automatic mesh generation technology. The component's response is then solved on the mesh elements to obtain the component's stress field or the stress field after stress equivalence.

[0013] Based on the SN curve fatigue criterion for high-cycle fatigue, the stress field or stress equivalent stress field is converted into an element life field, the element with the shortest life and its corresponding position coordinates are determined, and a pair of datasets is generated; the dataset includes sample input information and sample output information, and the sample output information includes the element life field and the position coordinates of the element with the shortest life.

[0014] The aforementioned pairs of datasets are used as training datasets and input into a deep learning model to establish the nonlinear mapping relationship model.

[0015] Furthermore, the input information and the information parameters also include defect geometric parameters.

[0016] Furthermore, the establishment of the component geometric model includes:

[0017] A geometric model of the component is established based on the geometric dimensions of the metal component, and the boundary constraint region and load application region of the component geometric model are determined based on the boundary condition parameters and the load application location parameters.

[0018] Furthermore, the step of using automatic mesh generation technology to divide the geometric model of the component into mesh elements, and then solving for the component response on the mesh elements to obtain the stress field of the component or the stress field after stress equivalence, includes:

[0019] The component geometric model is automatically discretized using automatic mesh generation technology to obtain a computational mesh composed of several mesh elements, and a mapping relationship is established between the boundary constraint region and the load application region to the mesh elements.

[0020] The component response is solved on the computational grid based on the load form parameters and the load magnitude parameters to obtain the stress field or stress equivalent stress field for high cycle fatigue calculation.

[0021] The automatic mesh generation technique is used to automatically discretize the geometric model of the component to obtain a computational mesh composed of several mesh elements, including:

[0022] Based on the component geometric model and preset mesh control parameters, the component geometric model is discretized using automatic mesh generation technology, and a computational mesh composed of several mesh units is automatically generated. The computational mesh is then adjusted and optimized according to the geometric boundary approximation error, unit size threshold, load application location neighborhood densification criterion, and mesh quality criterion, so that the mesh unit boundaries generated by the automatic mesh generation technology can fit the structural boundary and meet the computational accuracy requirements.

[0023] Furthermore, the component response is solved using finite element numerical calculation;

[0024] The finite element numerical calculation employs a numerical calculation method to solve for the discrete information of the grid element boundaries generated by the automatic grid generation technology, in order to obtain the element stress field with the stress amplitude or equivalent stress amplitude.

[0025] Furthermore, the SN curve fatigue criterion based on high-cycle fatigue converts the stress field or the stress field after stress equivalence into an element life field, including:

[0026] Based on the material's SN curve or its fitting parameters, the element stress amplitude is mapped to the element fatigue life, and an average stress correction is introduced to obtain the element life field.

[0027] Furthermore, the cell lifetime field is a set of lifetime scalars that correspond one-to-one with the grid cells generated by the automatic mesh generation technology;

[0028] The unit with the shortest lifespan and its corresponding position coordinates are obtained by searching for the unit with the shortest lifespan in the unit lifespan field and outputting the position coordinates of the center point of that unit.

[0029] Furthermore, establishing the nonlinear mapping relationship model includes:

[0030] A deep learning model is used to establish an explicit or implicit nonlinear mapping relationship model, so that the deep learning model outputs the cell lifetime field and the coordinates of the cell with the shortest lifetime after inputting the sample input information.

[0031] Furthermore, the deep learning model is a multi-task prediction model, which uses a weighted combination of the cell lifetime field prediction error and the shortest lifetime cell coordinate prediction error as the training objective function, while simultaneously optimizing the cell lifetime field prediction and the shortest lifetime cell coordinate prediction.

[0032] Secondly, the present invention also provides a rapid prediction system for high-cycle fatigue life and critical locations of metal components, comprising:

[0033] The acquisition module is used to acquire the input information of the metal component to be predicted;

[0034] The model processing module is used to input the input information into a pre-built nonlinear mapping relationship model to obtain the life field of the metal component and the position coordinates of the shortest life element. The input information includes the geometric dimensions of the metal component, boundary conditions, the location of the load application, the load type, the load magnitude, and material parameters. The life field of the metal component and the position coordinates of the shortest life element are the high-cycle fatigue life distribution and the critical location of the component, respectively.

[0035] The model processing module includes a model training unit, used for:

[0036] Several pairs of datasets are randomly generated, wherein the generation of any pair of datasets includes:

[0037] Obtain the information parameters of the target metal component and establish a geometric model of the component; the information parameters include the geometric dimension parameters, boundary condition parameters, load application location parameters, load form parameters, load magnitude parameters, and material parameters of the metal component;

[0038] The component's geometric model is divided into mesh elements using automatic mesh generation technology. The component's response is then solved on the mesh elements to obtain the component's stress field or the stress field after stress equivalence.

[0039] Based on the SN curve fatigue criterion for high-cycle fatigue, the stress field or stress equivalent stress field is converted into an element life field, the element with the shortest life and its corresponding position coordinates are determined, and a pair of datasets is generated; the dataset includes sample input information and sample output information, and the sample output information includes the element life field and the position coordinates of the element with the shortest life.

[0040] The aforementioned pairs of datasets are used as training datasets and input into a deep learning model to establish the nonlinear mapping relationship model.

[0041] The beneficial effects achieved by this invention are as follows:

[0042] The method provided by this invention utilizes a combination of parametric modeling, automatic mesh generation, and finite element numerical calculation to systematically construct a high-cycle fatigue life database covering different geometric dimensions, boundary conditions, load forms, and material parameters during the sample stage. This provides rich training data for the nonlinear mapping relationship model, thereby significantly improving the model's generalization ability. Compared to the traditional process of performing high-precision finite element fatigue analysis case by case, this invention, by pre-constructing the nonlinear mapping relationship model offline, only requires input of structural and working condition information during the prediction stage to quickly output the full-field life field and the shortest life location, greatly reducing computation time and resource consumption. It is suitable for rapid comparison and optimization design of multiple working conditions and multiple structural schemes. Attached Figure Description

[0043] Figure 1 This is a flowchart illustrating a rapid prediction method for high-cycle fatigue life and critical locations of metal components provided by the present invention.

[0044] Figure 2 This is a schematic diagram of the geometry of the notched metal plate provided in Example 4.

[0045] Figure 3 Input load signal for the notched metal plate provided in Example 4 .

[0046] Figure 4 The finite element calculation mesh for the notched metal plate structure provided in Example 4.

[0047] Figure 5 The diagram shows the deep learning network structure of the notched metal plate provided in Example 4.

[0048] Figure 6 The predicted life field and coordinate diagram of the most dangerous location of the metal plate with notches provided in Example 4. Detailed Implementation

[0049] The present invention will be further described below with reference to the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solution of the present invention, and should not be used to limit the scope of protection of the present invention.

[0050] Example 1, such as Figure 1 As shown, this invention discloses a rapid prediction method for high-cycle fatigue life and critical locations of metal components, comprising:

[0051] The input information of the metal component to be predicted is obtained and input into a pre-built nonlinear mapping relationship model. The corresponding component unit life field and the position coordinates of the shortest life unit are output. The component unit life field is used to characterize the full-field high-cycle fatigue life distribution of the component under a given working condition. The coordinates of the shortest life unit are used to characterize the location of the most dangerous part.

[0052] The input information includes: the component's geometric dimensions, boundary conditions, the location where the load is applied, the type of load, the magnitude of the load, and material parameters.

[0053] The construction of the nonlinear mapping relationship model includes:

[0054] The metal plate structure to be predicted is parametrically described to determine the component geometric parameters, boundary condition parameters, load application location parameters, load form parameters, load magnitude parameters, and material parameters, and to establish the value range of each parameter.

[0055] A geometric model of the component is established based on the component's geometric dimensions, and the boundary constraint region and load application region are determined based on the boundary condition parameters and the load application location parameters.

[0056] The component geometric model is automatically discretized using automatic mesh generation technology to obtain a computational mesh composed of several mesh elements, and a mapping relationship is established between the boundary constraint region and the load application region to the mesh elements.

[0057] The structural response is solved on the computational grid based on the load form and the load magnitude to obtain the stress amplitude or the element equivalent stress amplitude for high-cycle fatigue calculation.

[0058] Based on the SN curve fatigue criterion of high cycle fatigue, the element field of the stress amplitude or equivalent stress amplitude is converted into the structural element life field, and the coordinates of the element with the shortest life are determined, wherein the coordinates of the element with the shortest life are the geometric center coordinates of the element with the shortest life.

[0059] Multiple sets of sample input information that can characterize the structural working conditions are randomly generated in the parameter space. For each set of sample input information, the automatic mesh generation, structural response solution and element lifetime calculation based on SN curve are repeatedly executed to generate a pair of datasets. The pair of datasets includes sample input information and its corresponding sample output information. The sample output information includes the element lifetime field and the position coordinates of the element with the shortest lifetime. The process of generating datasets is repeated until multiple pairs of datasets that meet the conditions are generated.

[0060] Establish an explicit or implicit nonlinear mapping relationship model between sample input information and sample output information in the dataset to obtain the pre-constructed nonlinear mapping relationship model.

[0061] When discretizing the component geometric model using automatic mesh generation technology, the process includes: automatically generating a computational mesh composed of several mesh units based on the component geometric model and preset mesh control parameters; and adjusting and optimizing the computational mesh according to geometric boundary approximation error, unit size threshold, load application location neighborhood refinement criterion, and mesh quality criterion, so that the mesh unit boundaries generated by the automatic mesh generation technology can fit the component boundaries and meet the accuracy requirements of high-cycle fatigue calculation.

[0062] The high-cycle fatigue SN curve fatigue criterion includes: mapping the element stress amplitude to the element fatigue life based on the material SN curve or its fitting parameters, and introducing mean stress correction to obtain the element life field, thereby improving the accuracy of life prediction under mean stress or multiaxial stress conditions.

[0063] The cell lifetime field is a set of lifetime scalars that correspond one-to-one with the grid cells generated by the automatic mesh generation technology; the position coordinates of the cell with the shortest lifetime are obtained by searching for the cell with the shortest lifetime in the cell lifetime field and outputting the geometric center position coordinates of that cell.

[0064] The structural response is solved by finite element numerical calculation, which is based on the mesh element boundary discretization information generated by automatic mesh generation technology to obtain the element stress field with the stress amplitude or equivalent stress amplitude.

[0065] The step of establishing a nonlinear mapping model between sample input information and sample output information in the dataset includes: using a deep learning model to establish the explicit or implicit nonlinear mapping model, so that the deep learning model can simultaneously output the cell lifetime field and the location coordinates of the shortest-lived cell after receiving the sample input information. Preferably, the deep learning model is a multi-task prediction model, and the weighted combination of the cell lifetime field prediction error and the shortest-lived cell location coordinate prediction error is used as the training objective function to simultaneously optimize the performance of cell lifetime field prediction and shortest-lived cell coordinate prediction.

[0066] Example 2: This example introduces a rapid prediction system for high-cycle fatigue life and critical locations of metal components provided by the present invention, including:

[0067] The acquisition module is used to acquire the input information of the metal component to be predicted;

[0068] The model processing module is used to input the input information into a pre-built nonlinear mapping relationship model to obtain the life field of the metal component and the position coordinates of the shortest life element. The input information includes the geometric dimensions of the metal component, boundary conditions, the location of the load application, the load type, the load magnitude, and material parameters. The life field of the metal component and the position coordinates of the shortest life element are the high-cycle fatigue life distribution and the critical location of the component, respectively.

[0069] The model processing module includes a model training unit, used for:

[0070] Several pairs of datasets are randomly generated, wherein the generation of any pair of datasets includes:

[0071] Obtain the information parameters of the target metal component and establish a geometric model of the component; the information parameters include the geometric dimension parameters, boundary condition parameters, load application location parameters, load form parameters, load magnitude parameters, and material parameters of the metal component;

[0072] The component's geometric model is divided into mesh elements using automatic mesh generation technology. The component's response is then solved on the mesh elements to obtain the component's stress field or the stress field after stress equivalence.

[0073] Based on the SN curve fatigue criterion for high-cycle fatigue, the stress field or stress equivalent stress field is converted into an element life field, the element with the shortest life and its corresponding position coordinates are determined, and a pair of datasets is generated; the dataset includes sample input information and sample output information, and the sample output information includes the element life field and the position coordinates of the element with the shortest life.

[0074] The aforementioned pairs of datasets are used as training datasets and input into a deep learning model to establish the nonlinear mapping relationship model.

[0075] Example 3, based on the same inventive concept as other examples, provides a method for rapid prediction of high-cycle fatigue life and critical locations of metal components, including:

[0076] The input information of the component to be predicted is obtained and input into a pre-constructed nonlinear mapping model to obtain the life field of the component and the position coordinates of the shortest life element. The input information includes the component's geometric dimensions, boundary conditions, load application location, load type, load magnitude, and material parameters. The life field of the component and the position coordinates of the shortest life element are the high-cycle fatigue life distribution and critical location of the component, respectively.

[0077] The process of constructing the nonlinear mapping relationship model includes the following steps.

[0078] Step 1: Perform parametric modeling of the metal component to be predicted in a unified Cartesian coordinate system. The geometric dimensions of the component include its length. Component width and component thickness The component contains at least one geometric defect, described by the defect type and defect geometric parameters. The defect type includes at least one or a combination of circular hole notch, rectangular notch, and arc notch.

[0079] For a circular hole notch, use the center coordinates of the hole. and hole radius Description, i.e., defect geometric parameters ;

[0080] For rectangular gaps, use the coordinates of the rectangle's center. Length of the longer side and the length of the shorter side Description, i.e., defect geometric parameters ;

[0081] For the circular arc notch, use the coordinates of the center of the circle. ,radius and the arc angle Description, i.e., defect geometric parameters ;

[0082] Therefore, the geometric dimensional parameters and defect parameters of metal components can be uniformly represented as a set of geometric input parameters:

[0083] ;

[0084] Among them, subscript Indicates the first Group working condition sample.

[0085] Step 2: Based on the parameterized description given in Step 1, for any set of geometric input parameters Establish the corresponding geometric model of the metal component, and use the finite element method to mesh the geometric model to obtain a discrete numerical model;

[0086] Based on this, the boundary conditions, load types and material parameters in this invention are defined, and the input information for the deep learning model is constructed, including: boundary condition parameters, load application location and application range, cyclic load time history and material parameters.

[0087] The process of defining the boundary condition parameters includes:

[0088] Select the bottom edge of the plate as the boundary condition, and use a... Description of the The constraint form of this boundary under the combined working conditions; This indicates that the boundary is a fixed constraint; This indicates that the boundary is simply supported at both ends and free in the middle.

[0089] The process of defining the load application location and application range includes:

[0090] The top edge of the plate is selected as the load application point, and coordinates are used along the length of this boundary. Describe the range of load application; use the starting position. With interval length Description of the For ease of input, the two quantities mentioned above can be normalized to the location of the load application range for the group of working conditions as follows:

[0091] .

[0092] The process of defining the cyclic load time history includes:

[0093] No. The cyclic load for the group of working cases adopts a load function that varies with time. Description, the It can be a pre-defined standard time history, given here. ;

[0094] There are two types of loads: sinusoidal loads and cosine loads.

[0095] A sinusoidal load can be expressed as:

[0096] ;

[0097] In the formula, This represents the maximum pressure. Stress ratio, , This represents the minimum pressure value; Angular frequency, ; For time, This represents the number of loop iterations.

[0098] Cosine loads can be expressed as:

[0099] .

[0100] The process of defining the material parameters includes:

[0101] Material parameters and high-cycle fatigue S-N curve parameters are represented by a set of vectors. Characterization;

[0102] Therefore, the first The complete input parameter vector for the group of working condition samples can be written as:

[0103] .

[0104] Step 3: Within the parameter space defined in Steps 1 and 2, generate multiple sets of operating condition samples through random sampling or uniform sampling to obtain... Group input parameter vector:

[0105] ;

[0106] Each set of samples corresponds to a geometric dimension, notch type and defect location dimension, boundary conditions, load range and cyclic load time history;

[0107] For each group of samples Based on geometric parameters and defect parameters Establish the corresponding geometric model of the metal component and perform finite element mesh generation;

[0108] According to boundary condition parameters Apply constraints on the specified boundaries;

[0109] According to position parameters Apply cyclic loads within the corresponding interval of the load boundary. ;

[0110] The structural response is calculated using the finite element method to obtain the stress time history or stress amplitude information within the load cycle.

[0111] Step 4: Perform high-cycle fatigue life analysis on the stress response results obtained for each set of working conditions in Step 3, including:

[0112] Rainflow counting was performed on the stress time history of each unit to obtain the stress amplitude range and the corresponding number of cycles;

[0113] Using step 2 Given the S-N curve of the material, calculate the number of cycles it can withstand under different stress amplitudes;

[0114] The fatigue damage caused by stress cycles is accumulated using the linear damage accumulation theory to obtain the cumulative damage of each element. The cumulative damage Converted to unit fatigue life .

[0115] The first Fatigue life of all units under combined working conditions Arranged in order of element number, the lifetime field vector can be obtained:

[0116] ;

[0117] in, This represents the total number of elements in the geometric model.

[0118] Find the element with the shortest lifetime in the lifetime field, and denote the geometric center coordinates of this element as... As the first Coordinates of the element with the shortest lifespan under combined working conditions;

[0119] For the The group of samples and their corresponding output information are:

[0120] ;

[0121] By repeatedly executing steps 3 and 4, multiple sets of samples are generated and corresponding fatigue life calculations are performed until samples that meet the requirements are generated. For dataset .

[0122] Step 5: Combine the input and output information from Steps 3 and 4 to form the training dataset for the deep learning model;

[0123] Each set of training samples can be represented as:

[0124] .

[0125] Step 6: Establish using a deep learning model In the dataset and The explicit or implicit nonlinear mapping relationship between them enables the rapid generation of corresponding life field prediction results and shortest-life element coordinate prediction results, given any set of geometric dimensions, notch parameters, boundary condition parameters, and cyclic load parameters. It can output the corresponding .

[0126] Example 4, based on the same inventive concept as other examples, takes a rectangular metal plate structure with defects as an example to provide a rapid prediction method for high-cycle fatigue life and dangerous parts of metal components, including the following steps.

[0127] Step 1: As Figure 2 As shown, a geometric model of the metal plate to be predicted is established. A rectangular metal plate is selected with a length L of 100 mm, a width B of 200 mm, a thickness t of 1 mm, and an elastic modulus of 1 mm. Poisson's ratio mass density ;

[0128] The bottom edge of the plate is selected as the boundary condition, and the bottom edge of the plate is subject to a fixed constraint, denoted as . The plate has an arc-shaped notch, and the center coordinates of the arc are... The radius of the arc is 50 mm.

[0129] Step 2: As Figure 3 As shown, the cyclic load is randomly applied at any position on the top edge. In this embodiment, it is applied to the entire top edge as an example. mm, mm;

[0130] ;

[0131] Using a time-varying load function The description states that the total loading time is 0.2 seconds, the load type is sinusoidal, and the maximum pressure is... MPa, minimum pressure MPa, stress ratio ;

[0132] .

[0133] Step 3, as follows Figure 4 As shown, the geometric model is automatically discretized using quadtree automatic mesh generation in the proportional boundary finite element method to obtain a computational mesh composed of quadtree elements.

[0134] Step 4: Solve the structural response of the model to obtain the element field with stress amplitude or equivalent stress amplitude for high-cycle fatigue calculation;

[0135] Using the SN curve fatigue criterion of high-cycle fatigue, the element field of the stress amplitude or equivalent stress amplitude is converted into the structural element lifetime field, and the coordinates of the element with the shortest lifetime are determined, where the reference lifetime is... Secondary, reference stress amplitude Basquin slope .

[0136] Through steps 1 to 4, the input information for deep learning can be obtained. and output information .

[0137] Step 5: Repeat steps 1 to 4 to generate results that meet the conditions. For dataset .

[0138] Step 6, as follows Figure 5 As shown, a deep learning network structure model is established, and deep learning techniques are used to build it. In the dataset and The implicit nonlinear mapping relationship between them allows for rapid prediction of the fatigue life of a metal structure under evaluation. When this prediction is performed, the corresponding geometric parameters, defect parameters, boundary conditions, cyclic loads, and material parameters are used as input vectors. These vectors are then fed into a trained deep learning model to quickly obtain the predicted life field and the coordinates of the shortest-life element. Input a deep learning model, and it can output the corresponding... .

[0139] Step 7: Input information of the metal structure to be predicted The input is repeatedly fed into the deep learning model, and the training is repeated 400 times, resulting in 400 different outputs. The mean of these 400 samples is calculated, and this mean is taken as the final predicted value. The final predicted lifetime field and the coordinates of the most dangerous location are as follows: Figure 6 As shown.

[0140] Example 5, based on the same inventive concept as other examples, provides a rapid prediction system for high-cycle fatigue life and critical locations of metal components, comprising:

[0141] The acquisition module is used to acquire the plate's geometric dimensions, defect parameters, boundary condition parameters, cyclic load parameters, and material parameters to form the input vector. ;

[0142] The model processing module is used to process the input vector. Input is given to a pre-built nonlinear mapping model, and the corresponding output is obtained. , This refers to the lifetime field prediction results and the coordinate prediction results of the shortest-life unit, which characterize the structure.

[0143] The model processing module includes a model training unit, used for:

[0144] Based on the input information, a structural geometric model is established, and an automatic mesh generation technique is used to generate a computational mesh, establishing a mapping relationship between boundary conditions and loads to mesh elements;

[0145] Generate multiple sets of sample input information within the parameter space. And input information for each group of samples. By performing automatic mesh generation technology and finite element numerical calculations, the element field with element stress amplitude or equivalent stress amplitude is obtained. Based on the SN curve fatigue criterion for high-cycle fatigue, the corresponding element lifetime field and the coordinates of the element with the shortest lifetime are obtained. Generate a pair of datasets ;

[0146] By repeatedly executing the process of generating datasets, multiple pairs of datasets that meet the conditions are generated;

[0147] Establish an explicit or implicit nonlinear mapping model between sample input information and sample output information in the dataset;

[0148] After inputting the input information, the nonlinear mapping relationship model is invoked to output the life field of the structural unit and the coordinates of the unit with the shortest life.

[0149] In summary, the prediction method of this invention can be applied to the prediction of the lifespan of metal components in mechanical engineering and hydraulic engineering. This invention utilizes a combination of parametric modeling, automatic mesh generation, and finite element numerical calculation to systematically construct a high-cycle fatigue life database covering different geometric dimensions, boundary conditions, load forms, and material parameters during the sample stage. This provides rich training data for the nonlinear mapping relationship model, thereby significantly improving the model's generalization ability. Compared to the traditional process of performing high-precision finite element fatigue analysis case by case, this invention, by pre-constructing the nonlinear mapping relationship model offline, only requires input of structural and working condition information during the prediction stage to quickly output the full-field life field and the shortest lifespan location, greatly reducing computation time and resource consumption. It is suitable for rapid comparison and optimization design of multiple working conditions and multiple structural schemes.

[0150] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0151] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0152] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0153] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0154] The embodiments of the present invention have been described above with reference to the accompanying drawings. However, the present invention is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms under the guidance of the present invention without departing from the spirit and scope of the claims. All of these forms are within the protection scope of the present invention.

Claims

1. A rapid prediction method for high-cycle fatigue life and critical locations of metal components, characterized in that, include: The input information of the metal component to be predicted is obtained and input into a pre-constructed nonlinear mapping model to obtain the life field and the position coordinates of the shortest life element of the metal component. The input information includes the geometric dimensions of the metal component, boundary conditions, load application location, load type, load magnitude and material parameters. The life field and the position coordinates of the shortest life element of the metal component are respectively the high-cycle fatigue life distribution and the critical location of the metal component. The construction process of the nonlinear mapping relationship model includes: Several pairs of datasets are randomly generated, wherein the generation of any pair of datasets includes: Obtain the information parameters of the target metal component and establish a geometric model of the component; the information parameters include the geometric dimension parameters, boundary condition parameters, load application location parameters, load form parameters, load magnitude parameters, and material parameters of the metal component; The component's geometric model is divided into mesh elements using automatic mesh generation technology. The component's response is then solved on the mesh elements to obtain the component's stress field or the stress field after stress equivalence. Based on the SN curve fatigue criterion for high-cycle fatigue, the stress field or stress equivalent stress field is converted into an element life field, the element with the shortest life and its corresponding position coordinates are determined, and a pair of datasets is generated; the dataset includes sample input information and sample output information, and the sample output information includes the element life field and the position coordinates of the element with the shortest life. The aforementioned pairs of datasets are used as training datasets and input into a deep learning model to establish the nonlinear mapping relationship model.

2. The method for rapid prediction of high-cycle fatigue life and critical locations of metal components according to claim 1, characterized in that, The input information and the information parameters also include defect geometric parameters.

3. The method for rapid prediction of high-cycle fatigue life and critical locations of metal components according to claim 1, characterized in that, The establishment of the component geometric model includes: A geometric model of the component is established based on the geometric dimensions of the metal component, and the boundary constraint region and load application region of the component geometric model are determined based on the boundary condition parameters and the load application location parameters.

4. The rapid prediction method for high-cycle fatigue life and critical locations of metal components according to claim 1, characterized in that, The process of using automatic mesh generation technology to divide the geometric model of the component into mesh elements, and then solving for the component response on the mesh elements to obtain the stress field of the component or the stress field after stress equivalence includes: The component geometric model is automatically discretized using automatic mesh generation technology to obtain a computational mesh composed of several mesh elements, and a mapping relationship is established between the boundary constraint region and the load application region to the mesh elements. The component response is solved on the computational grid based on the load form parameters and the load magnitude parameters to obtain the stress field or stress equivalent stress field for high cycle fatigue calculation. The automatic mesh generation technique is used to automatically discretize the geometric model of the component to obtain a computational mesh composed of several mesh elements, including: Based on the component geometric model and preset mesh control parameters, the component geometric model is discretized using automatic mesh generation technology, and a computational mesh composed of several mesh units is automatically generated. The computational mesh is then adjusted and optimized according to the geometric boundary approximation error, unit size threshold, load application location neighborhood densification criterion, and mesh quality criterion, so that the mesh unit boundaries generated by the automatic mesh generation technology can fit the structural boundary and meet the computational accuracy requirements.

5. The rapid prediction method for high-cycle fatigue life and critical locations of metal components according to claim 4, characterized in that, The component response is solved by finite element numerical calculation; The finite element numerical calculation employs a numerical calculation method to solve for the discrete information of the grid element boundaries generated by the automatic grid generation technology, in order to obtain the element stress field with the stress amplitude or equivalent stress amplitude.

6. The rapid prediction method for high-cycle fatigue life and critical locations of metal components according to claim 1, characterized in that, The SN curve fatigue criterion based on high-cycle fatigue converts the stress field or the stress field after stress equivalence into an element life field, including: Based on the material's SN curve or its fitting parameters, the element stress amplitude is mapped to the element fatigue life, and an average stress correction is introduced to obtain the element life field.

7. The method for rapid prediction of high-cycle fatigue life and critical locations of metal components according to claim 1, characterized in that, The cell lifetime field is a set of lifetime scalars that correspond one-to-one with the grid cells generated by the automatic grid generation technology. The unit with the shortest lifespan and its corresponding position coordinates are obtained by searching for the unit with the shortest lifespan in the unit lifespan field and outputting the position coordinates of the center point of that unit.

8. The method for rapid prediction of high-cycle fatigue life and critical locations of metal components according to claim 1, characterized in that, The establishment of the nonlinear mapping relationship model includes: A deep learning model is used to establish an explicit or implicit nonlinear mapping relationship model, so that the deep learning model outputs the cell lifetime field and the coordinates of the cell with the shortest lifetime after inputting the sample input information.

9. The method for rapid prediction of high-cycle fatigue life and critical locations of metal components according to claim 8, characterized in that, The deep learning model is a multi-task prediction model, which uses a weighted combination of the cell lifetime field prediction error and the shortest lifetime cell coordinate prediction error as the training objective function, and simultaneously optimizes the cell lifetime field prediction and the shortest lifetime cell coordinate prediction.

10. A rapid prediction system for high-cycle fatigue life and critical locations of metal components, characterized in that, include: The acquisition module is used to acquire the input information of the metal component to be predicted; The model processing module is used to input the input information into a pre-built nonlinear mapping relationship model to obtain the life field of the metal component and the position coordinates of the shortest life element. The input information includes the geometric dimensions of the metal component, boundary conditions, the location of the load application, the load type, the load magnitude, and material parameters. The life field of the metal component and the position coordinates of the shortest life element are the high-cycle fatigue life distribution and the critical location of the component, respectively. The model processing module includes a model training unit, used for: Several pairs of datasets are randomly generated, wherein the generation of any pair of datasets includes: Obtain the information parameters of the target metal component and establish a geometric model of the component; the information parameters include the geometric dimension parameters, boundary condition parameters, load application location parameters, load form parameters, load magnitude parameters, and material parameters of the metal component; The component's geometric model is divided into mesh elements using automatic mesh generation technology. The component's response is then solved on the mesh elements to obtain the component's stress field or the stress field after stress equivalence. Based on the SN curve fatigue criterion for high-cycle fatigue, the stress field or stress equivalent stress field is converted into an element life field, the element with the shortest life and its corresponding position coordinates are determined, and a pair of datasets is generated; the dataset includes sample input information and sample output information, and the sample output information includes the element life field and the position coordinates of the element with the shortest life. The aforementioned pairs of datasets are used as training datasets and input into a deep learning model to establish the nonlinear mapping relationship model.

Images