Ceramic bearing roller crack control method, system and device based on artificial intelligence

Abstract

The present invention discloses a method, system and device for controlling cracks of ceramic bearing rollers based on artificial intelligence. The method includes: obtaining material physical parameters of ceramic bearing rollers; constructing a three-dimensional microstructure digital twin model to obtain an initial finite element model of the ceramic bearing rollers; applying mechanical boundary conditions and thermal boundary conditions to the initial finite element model to obtain a comprehensive finite element model to obtain thermal-mechanical coupling simulation data; obtaining initial crack parameters based on a crack initiation model to obtain crack propagation simulation information; obtaining optimal processing parameters based on crack propagation simulation information and equipment processing parameters; obtaining a crack risk prediction value for the current period based on a crack risk prediction model, obtaining and adjusting the actual optimal processing parameters for the current period, and realizing control of cracks of ceramic bearing rollers. The present invention effectively reduces the crack risk of ceramic bearing rollers during processing and improves the processing quality and production efficiency of ceramic bearing rollers.

Description

Technical Field

[0001] The present invention relates to the field of intelligent processing, and in particular to an artificial intelligence-based ceramic bearing roller crack control method, system and device. Background Art

[0002] Ceramic bearing rollers have been widely used in high-precision fields such as aerospace and mechanical manufacturing due to their excellent wear resistance, high-temperature stability and corrosion resistance. However, the brittleness of ceramic materials makes it very easy for ceramic bearing rollers to crack during processing. The formation and expansion of cracks will not only significantly affect product quality, but also reduce the service life of ceramic bearing rollers. At present, the existing crack control technology mainly relies on post-detection and manual adjustment of processing parameters, which makes it difficult to achieve real-time and accurate crack monitoring and control in a dynamic processing environment. In addition, the existing methods lack effective analysis of the interaction between temperature field and stress field during processing, and cannot predict the risk of crack extension. In order to improve the processing quality of ceramic rollers, there is an urgent need for an intelligent crack control method based on multi-parameter regulation, which can dynamically adjust key parameters during processing, monitor crack extension in real time, and thus improve processing accuracy and production efficiency. Summary of the invention

[0003] In view of the defects in the prior art, the present invention provides a method, system and device for controlling cracks in ceramic bearing rollers based on artificial intelligence.

[0004] In order to solve the above technical problems, the present invention is solved by the following technical solutions:

[0005] A method for controlling cracks in ceramic bearing rollers based on artificial intelligence comprises the following steps:

[0006] Obtain the material physical parameters of ceramic bearing rollers;

[0007] Obtain a 2D microscopic image sequence of the ceramic bearing roller and perform preprocessing to build a 3D microstructure digital twin model. Perform finite element analysis on the 3D microstructure digital twin model and combine it with the material physical parameters to obtain the initial finite element model of the ceramic bearing roller.

[0008] Apply mechanical boundary conditions and thermal boundary conditions to the initial finite element model to obtain a comprehensive finite element model, and combine the physical parameters of the material to obtain thermal-mechanical coupling simulation data, wherein the thermal-mechanical coupling simulation data at least includes a final temperature field, a final stress field, and a set of locations of high crack risk areas;

[0009] A crack initiation pre-training model is constructed and trained through thermal-mechanical coupling simulation data to obtain a crack initiation model, and then obtain initial crack parameters, wherein the initial crack parameters at least include an initial crack position, an initial crack direction, and an initial crack phenotype parameter;

[0010] Obtain crack propagation simulation information through initial crack parameters, comprehensive finite element model and material physical parameters, wherein the crack propagation simulation information at least includes crack propagation speed, maximum stress and maximum temperature;

[0011] Based on the crack propagation simulation information and equipment processing parameters, an optimization model is constructed and the equipment processing parameters are optimized to obtain the optimal processing parameters of the equipment;

[0012] Based on the crack risk prediction model, the processing parameters of the current period are inferred to obtain the crack risk prediction value of the current period. Based on the crack risk prediction value of the current period, the optimal processing parameters of the current period and the processing parameters of the current period, the actual optimal processing parameters of the current period are adjusted to achieve control of the cracks in the ceramic bearing roller.

[0013] As an implementable method, the two-dimensional microscopic image sequence of the ceramic bearing roller is obtained and preprocessed, and then a three-dimensional microstructure digital twin model is constructed, and a finite element analysis is performed on the three-dimensional microstructure digital twin model and combined with material physical parameters to obtain an initial finite element model of the ceramic bearing roller, including the following steps:

[0014] The ceramic bearing roller is scanned layer by layer to obtain a sequence of two-dimensional microscopic images of the ceramic bearing roller at different depths;

[0015] Preprocessing the two-dimensional microscopic image sequence to obtain a preprocessed two-dimensional microscopic image sequence, wherein the preprocessing includes one or more of denoising, convolution operation, equalization, segmentation, removing isolated pixels and filling holes;

[0016] Based on the preprocessed two-dimensional microscopic image sequence and combined with three-dimensional reconstruction technology, the corresponding three-dimensional microstructure digital twin model is obtained;

[0017] Finite element analysis is performed on the three-dimensional microstructure digital twin model to obtain an initial finite element model of the ceramic bearing roller, wherein the finite element analysis at least includes meshing and material property assignment, different regions are formed through meshing, and corresponding material physical parameters are assigned to different regions.

[0018] As an implementable method, applying mechanical boundary conditions and thermal boundary conditions to the initial finite element model to obtain a comprehensive finite element model, combining material physical parameters to obtain thermal-mechanical coupling simulation data, includes the following steps:

[0019] Applying mechanical boundary conditions to the initial finite element model to achieve displacement constraints, thereby obtaining a first finite element model;

[0020] Applying thermal boundary conditions to the first finite element model to obtain a comprehensive finite element model, wherein the thermal boundary conditions include temperature boundary conditions and heat flow boundary conditions;

[0021] Based on the comprehensive finite element model and material physical parameters, the thermal-mechanical coupling simulation data is obtained. The specific process includes:

[0022] Constructing control equations, wherein the control equations include at least a mechanical equilibrium equation, a heat conduction equation, and a constitutive relationship equation, and solving the control equations by finite element discretization and time discretization in combination with a comprehensive finite element model to obtain a final temperature field, a final stress field, and a set of locations of high crack risk areas;

[0023] The mechanical equilibrium equation, heat conduction equation and constitutive relationship equation are expressed as follows:

[0024]

[0025]

[0026]

[0027] in, represents stress, represents body density, Represents the density of ceramic materials, represents displacement, Indicates time, represents the specific heat capacity of ceramic materials, Indicates temperature, represents the thermal conductivity of ceramic materials, represents the heat source term, represents the thermal conductivity term, represents the elastic stiffness matrix, represents the total strain, represents the thermal strain caused by temperature change, represents the matrix transpose, represents the thermal expansion coefficient of ceramic materials, represents the reference temperature, Represents the identity matrix.

[0028] As an implementable method, obtaining the high crack risk area location set includes the following steps:

[0029] Based on the final stress field, the maximum principal stress in each region is obtained;

[0030] The areas where the maximum principal stress in the comprehensive finite element model is not less than the stress threshold are identified, and the location set of high crack risk areas is obtained, which is expressed as follows:

[0031]

[0032] in, represents the location set of high crack risk areas, represents the maximum principal stress, represents the stress threshold, Indicates a location point.

[0033] As an implementable method, the method of constructing a crack initiation pre-training model and training it through thermal-mechanical coupling simulation data to obtain a crack initiation model and then obtain initial crack parameters includes the following steps:

[0034] Construct a crack initiation pre-training model based on a three-dimensional space artificial intelligence model;

[0035] The crack initiation pre-training model is trained and tested through thermal-mechanical coupling simulation data to obtain a crack initiation model;

[0036] Input the thermal-mechanical coupling simulation data to be tested into the crack initiation model to obtain a set of predicted high crack risk area locations;

[0037] Based on the predicted high crack risk area location set, initial crack parameters are obtained, wherein the initial crack parameters at least include initial crack location, initial crack direction, and initial crack phenotype parameters.

[0038] As an implementable method, it also includes data preprocessing of the thermal-mechanical coupling simulation data, specifically:

[0039] Based on the final stress field and the final temperature field, the initial multi-channel three-dimensional input tensor is constructed, which is expressed as follows:

[0040]

[0041] in, represents the initial multi-channel three-dimensional input tensor, The final stress field The normal stress component of represents the shear stress component of the final stress field, represents the final temperature field;

[0042] Normalizing the initial multi-channel three-dimensional input tensor and reducing the three-dimensional data size to obtain the final multi-channel three-dimensional input tensor;

[0043] Based on the location set of high crack risk areas, a crack risk label tensor is constructed, which is expressed as follows:

[0044]

[0045] in, represents the crack risk label tensor, represents the location set of high crack risk areas, Indicates a location point;

[0046] A sample set is formed based on the final multi-channel three-dimensional input tensor and the crack risk label tensor.

[0047] As an implementable method, obtaining initial crack parameters based on the predicted high crack risk area location set includes the following steps:

[0048] Calculate the centroid position of the predicted high crack risk area location set to obtain the initial crack position, which is expressed as follows:

[0049]

[0050] in, The centroid of the predicted high crack risk area location set is the initial crack location. represents the predicted high crack risk area location set, The number of voxels representing the set of predicted high crack risk area locations;

[0051] Fitting the initial crack shape according to the spatial distribution of the predicted high crack risk area location set to obtain initial crack phenotype parameters, wherein the initial crack phenotype parameters include initial crack shape and initial crack size;

[0052] The maximum principal stress of the predicted high crack risk area location set is calculated, and then the average value of the maximum principal stress direction is obtained. The average value of the maximum principal stress direction is normalized to a unit vector, and the unit vector is the initial crack direction. The average value of the maximum principal stress direction is expressed as follows:

[0053]

[0054] in, represents the average value of the maximum principal stress direction of the predicted high crack risk area location set, is the eigenvector of the maximum principal stress of the predicted high crack risk area location set.

[0055] As an implementable method, the method of obtaining crack propagation simulation information through initial crack parameters, comprehensive finite element model and material physical parameters includes the following steps:

[0056] The initial crack position is used as a reference system, and the initial crack phenotype parameters are matched to the comprehensive finite element model. Then, the crack surface normal vector of the comprehensive finite element model is matched to the initial crack direction and the mesh of the initial crack area is refined to update the comprehensive finite element model. The crack propagation condition, crack propagation direction and crack propagation increment are defined to obtain the crack propagation finite element model.

[0057] Based on the crack propagation finite element model, the control equation is iteratively solved through spatial discretization and time discretization. Based on the data of each iteration in the solution process, the crack propagation simulation information is obtained. The solution process is the crack propagation simulation process, which specifically includes:

[0058] When the crack propagation conditions are met, crack propagation is performed to obtain the current crack propagation speed, crack propagation increment, crack propagation direction, stress field and temperature field, and then the crack propagation finite element model is updated, wherein the crack propagation increment is obtained based on the crack propagation speed;

[0059] Based on the updated crack propagation finite element model, the control equation is re-solved until the solution termination condition is met, and then the crack propagation simulation process is terminated;

[0060] Based on the stress field and temperature field of each iteration, the maximum stress and maximum temperature are obtained.

[0061] As an implementable embodiment, the crack extension condition includes performing crack extension if the current crack energy release rate is not less than the energy release threshold, and the energy release rate is expressed as follows:

[0062]

[0063] in, represents the current energy release rate, represents the equivalent elastic modulus of the ceramic bearing roller, represents the stress intensity factor of the opening mode of the previous iteration, represents the slip mode stress intensity factor of the previous iteration, represents the stress intensity factor of the tearing mode of the previous iteration, which are expressed as follows:

[0064]

[0065]

[0066]

[0067] in, represents the maximum principal stress in the normal direction of the current crack tip, represents the crack length of the previous iteration, represents the tangential shear stress of the stress tensor at the current crack tip, The normal shear stress represents the stress tensor at the current crack tip;

[0068] The crack propagation direction is expressed as follows:

[0069]

[0070] in, represents the current crack propagation direction, represents the crack growth direction of the previous iteration, represents the crack extension angle of the previous iteration, ;

[0071] The crack growth rate and the crack growth increment are expressed as follows:

[0072]

[0073]

[0074] in, represents the current crack growth rate, represents the crack growth rate constant, Indicates the critical energy release rate of the material of the ceramic bearing roller, i.e. the energy release threshold. represents the crack growth index, represents the current crack growth increment, Indicates the time difference.

[0075] As an implementable method, the optimization model is constructed based on the crack propagation simulation information and the equipment processing parameters, and the equipment processing parameters are optimized to obtain the optimal processing parameters of the equipment, including the following steps:

[0076] Constructing an optimization model, wherein the optimization model at least includes an optimization variable model, an optimization target model and a constraint condition model. The optimization variable model is constructed based on equipment processing parameters. The equipment processing parameters are optimization variables that at least include cutting speed, feed rate, cutting depth and coolant flow rate. The optimization target model is constructed based on a first crack risk value and a processing cost value. The first crack risk value is obtained based on crack extension simulation information. The processing cost value at least includes a processing time cost value, a tool wear cost value and an energy consumption cost value. The constraint condition model includes an equipment processing parameter constraint condition model and a processing quality constraint condition model. The processing quality constraint condition model is constructed based on surface roughness, processing dimensional accuracy and processing tool wear.

[0077] The optimization model is preliminarily solved by genetic algorithm, and the optimal population is obtained as the preliminary optimization processing parameters. Among them, the population to be optimized in the genetic algorithm is the equipment processing parameters, the fitness of the population is obtained by optimizing the target model, and the encoding of the population is controlled by the constraint condition model;

[0078] The optimization model is globally solved by the particle swarm algorithm, and the global optimal position of the particle swarm is the optimal processing parameter of the equipment. The initial position of the particle is the preliminary optimized processing parameter, the fitness of the particle is obtained by optimizing the target model, and the position of the particle is adjusted by the constraint condition model.

[0079] The optimization variable model, the optimization target model, the equipment processing parameter constraint model, the processing quality constraint model, the first crack risk value and the processing cost value are respectively expressed as follows:

[0080]

[0081]

[0082]

[0083]

[0084]

[0085]

[0086] in, represents the optimization variable, i.e., the equipment processing parameters, Indicates the cutting speed, Indicates the feed rate, Indicates the cutting depth, Indicates the coolant flow rate, represents the optimization goal, represents the weight coefficient, and , represents the first crack risk value, Indicates the processing cost value, Indicates the surface roughness of ceramic bearing rollers. Indicates the machining dimensional accuracy required for ceramic bearing rollers. Indicates the wear of the machining tool, Indicates the maximum and minimum values ​​of the corresponding parameters, represents the maximum stress, is the crack growth rate, Indicates the maximum temperature, Both represent weight coefficients, represents the processing time cost value, , represents the unit time cost, Indicates the length of the ceramic bearing roller, represents the tool wear cost value, , represents the unit tool wear cost value, represents the energy consumption cost value, , represents the unit energy cost, Indicates energy consumption.

[0087] As an implementable method, the method infers the processing parameters of the current period based on the crack risk prediction model to obtain the crack risk prediction value of the current period, and adjusts the actual optimal processing parameters of the current period based on the crack risk prediction value of the current period, the optimal processing parameters of the current period and the processing parameters of the current period to achieve the control of the cracks of the ceramic bearing roller, including the following steps:

[0088] Constructing a crack risk prediction pre-training model, and training the crack risk prediction pre-training model based on a historical first processing parameter of the equipment to obtain a crack risk prediction model, wherein the first processing parameter at least includes cutting force data, vibration data, temperature data, and a crack risk value;

[0089] The first processing parameter of the equipment in the current period is inferred by the crack risk prediction model to obtain the crack risk prediction value in the current period;

[0090] Based on the crack risk prediction value of the current period, the optimal processing parameters of the current period and the processing parameters of the current period, the adjusted processing parameters of the current period are obtained, specifically including: obtaining the risk error based on the crack risk prediction value of the current period and the preset target crack risk level, obtaining the adjustment amount of the theoretical processing parameters of the current period based on the risk error, the optimal processing parameters of the current period and the processing parameters of the current period, and then obtaining the theoretical optimal processing parameters of the current period, limiting the theoretical optimal processing parameters of the current period within a preset range, and obtaining the adjusted processing parameters of the current period;

[0091] Adjusting the actual optimal processing parameters of the current period based on the adjusted processing parameters of the current period to achieve the control of the cracks of the ceramic bearing roller;

[0092] Among them, the risk error, the theoretical processing parameter adjustment amount in the current period, and the adjusted processing parameters in the current period are expressed as follows:

[0093]

[0094]

[0095]

[0096] in, represents the risk error, Indicates the preset target crack risk level, Indicates the crack risk prediction value of the current period, Indicates the theoretical processing parameter adjustment amount in the current period. represents the control gain, represents the weight coefficient, Indicates the optimal processing parameters for the current period, Indicates the processing parameters of the current period. Indicates the processing parameters adjusted in the current period. Indicates the theoretical optimal processing parameters for the current period, and Indicates the minimum and maximum values ​​preset for the processing parameters. Indicates that the target parameter is limited between the preset minimum and maximum values.

[0097] A ceramic bearing roller crack control system based on artificial intelligence, including a material parameter acquisition module, a finite element model acquisition module, a thermal data acquisition module, a crack parameter acquisition module, a crack extension information acquisition module, an optimal processing parameter acquisition module and an actual optimal processing parameter acquisition module;

[0098] The material parameter acquisition module is used to acquire the material physical parameters of the ceramic bearing roller;

[0099] The finite element model acquisition module is used to acquire a two-dimensional microscopic image sequence of the ceramic bearing roller and perform preprocessing, thereby constructing a three-dimensional microstructure digital twin model, performing finite element analysis on the three-dimensional microstructure digital twin model and combining it with material physical parameters to obtain an initial finite element model of the ceramic bearing roller;

[0100] The thermal data acquisition module is used to apply mechanical boundary conditions and thermal boundary conditions to the initial finite element model to obtain a comprehensive finite element model, and to obtain thermal-mechanical coupling simulation data in combination with material physical parameters, wherein the thermal-mechanical coupling simulation data at least includes a final temperature field, a final stress field, and a set of locations of high crack risk areas;

[0101] The crack parameter acquisition module is used to construct a crack initiation pre-training model and train it through thermal-mechanical coupling simulation data to obtain a crack initiation model, and then obtain initial crack parameters, wherein the initial crack parameters at least include initial crack position, initial crack direction and initial crack phenotype parameters;

[0102] The crack extension information acquisition module is used to obtain crack extension simulation information through initial crack parameters, comprehensive finite element model and material physical parameters, wherein the crack extension simulation information at least includes crack extension speed, maximum stress and maximum temperature;

[0103] The optimal processing parameter acquisition module is used to construct an optimization model and optimize the equipment processing parameters based on the crack propagation simulation information and the equipment processing parameters to obtain the optimal processing parameters of the equipment;

[0104] The actual optimal processing parameter acquisition module is used to infer the processing parameters of the current period based on the crack risk prediction model to obtain the crack risk prediction value of the current period, and adjust the actual optimal processing parameters of the current period based on the crack risk prediction value of the current period, the optimal processing parameters of the current period and the processing parameters of the current period to achieve control of the cracks of the ceramic bearing roller.

[0105] A computer program product comprises computer instructions, which, when executed by a processor, implement the following method:

[0106] Obtain the material physical parameters of ceramic bearing rollers;

[0107] Obtain a 2D microscopic image sequence of the ceramic bearing roller and perform preprocessing to build a 3D microstructure digital twin model. Perform finite element analysis on the 3D microstructure digital twin model and combine it with the material physical parameters to obtain the initial finite element model of the ceramic bearing roller.

[0108] Apply mechanical boundary conditions and thermal boundary conditions to the initial finite element model to obtain a comprehensive finite element model, and combine the physical parameters of the material to obtain thermal-mechanical coupling simulation data, wherein the thermal-mechanical coupling simulation data at least includes a final temperature field, a final stress field, and a set of locations of high crack risk areas;

[0109] A crack initiation pre-training model is constructed and trained through thermal-mechanical coupling simulation data to obtain a crack initiation model, and then obtain initial crack parameters, wherein the initial crack parameters at least include an initial crack position, an initial crack direction, and an initial crack phenotype parameter;

[0110] Obtain crack propagation simulation information through initial crack parameters, comprehensive finite element model and material physical parameters, wherein the crack propagation simulation information at least includes crack propagation speed, maximum stress and maximum temperature;

[0111] Based on the crack propagation simulation information and equipment processing parameters, an optimization model is constructed and the equipment processing parameters are optimized to obtain the optimal processing parameters of the equipment;

[0112] Based on the crack risk prediction model, the processing parameters of the current period are inferred to obtain the crack risk prediction value of the current period. Based on the crack risk prediction value of the current period, the optimal processing parameters of the current period and the processing parameters of the current period, the actual optimal processing parameters of the current period are adjusted to achieve control of the cracks in the ceramic bearing roller.

[0113] A ceramic bearing roller crack control device based on artificial intelligence includes a memory, a processor, and a computer program stored in the memory and running on the processor. When the processor executes the computer program, the following method is implemented:

[0114] Obtain the material physical parameters of ceramic bearing rollers;

[0115] Obtain a 2D microscopic image sequence of the ceramic bearing roller and perform preprocessing to build a 3D microstructure digital twin model. Perform finite element analysis on the 3D microstructure digital twin model and combine it with the material physical parameters to obtain the initial finite element model of the ceramic bearing roller.

[0116] Apply mechanical boundary conditions and thermal boundary conditions to the initial finite element model to obtain a comprehensive finite element model, and combine the physical parameters of the material to obtain thermal-mechanical coupling simulation data, wherein the thermal-mechanical coupling simulation data at least includes a final temperature field, a final stress field, and a set of locations of high crack risk areas;

[0117] A crack initiation pre-training model is constructed and trained through thermal-mechanical coupling simulation data to obtain a crack initiation model, and then obtain initial crack parameters, wherein the initial crack parameters at least include an initial crack position, an initial crack direction, and an initial crack phenotype parameter;

[0118] Obtain crack propagation simulation information through initial crack parameters, comprehensive finite element model and material physical parameters, wherein the crack propagation simulation information at least includes crack propagation speed, maximum stress and maximum temperature;

[0119] Based on the crack propagation simulation information and equipment processing parameters, an optimization model is constructed and the equipment processing parameters are optimized to obtain the optimal processing parameters of the equipment;

[0120] Based on the crack risk prediction model, the processing parameters of the current period are inferred to obtain the crack risk prediction value of the current period. Based on the crack risk prediction value of the current period, the optimal processing parameters of the current period and the processing parameters of the current period, the actual optimal processing parameters of the current period are adjusted to achieve control of the cracks in the ceramic bearing roller.

[0121] The present invention has significant technical effects due to the adoption of the above technical solution:

[0122] The present invention flexibly uses a variety of artificial intelligence models, obtains data related to cracks in ceramic bearing rollers through intelligent methods, effectively analyzes the interaction between temperature field and stress field during processing, and then predicts the risk of crack extension, realizes real-time and precise control of parameters during the processing of ceramic bearing rollers, and monitors crack extension in real time. By intelligently controlling the cracks generated during the processing, the risk of cracks in ceramic bearing rollers during the processing is effectively reduced, and the processing quality and production efficiency of ceramic bearing rollers are improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0123] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0124] Figure 1 It is a schematic flow diagram of the method of the present invention;

[0125] Figure 2 It is an overall schematic diagram of the system of the present invention. DETAILED DESCRIPTION

[0126] The present invention is further described in detail below in conjunction with the embodiments, which are explanations of the present invention but are not limited to the following embodiments. In the absence of conflict, the features in the following embodiments can be combined with each other.

[0127] Embodiment 1:

[0128] A method for controlling cracks in ceramic bearing rollers based on artificial intelligence, such as Figure 1 As shown, the following steps are included:

[0129] S100: obtaining material physical parameters of ceramic bearing rollers;

[0130] S200: obtaining a two-dimensional microscopic image sequence of a ceramic bearing roller and performing preprocessing, thereby constructing a three-dimensional microstructure digital twin model, performing finite element analysis on the three-dimensional microstructure digital twin model and combining it with material physical parameters to obtain an initial finite element model of the ceramic bearing roller;

[0131] S300: applying mechanical boundary conditions and thermal boundary conditions to the initial finite element model to obtain a comprehensive finite element model, and combining material physical parameters to obtain thermal-mechanical coupling simulation data, wherein the thermal-mechanical coupling simulation data at least includes a final temperature field, a final stress field, and a set of locations of high crack risk areas;

[0132] S400: constructing a crack initiation pre-training model and training it through thermal-mechanical coupling simulation data to obtain a crack initiation model, and then obtain initial crack parameters, wherein the initial crack parameters at least include an initial crack position, an initial crack direction, and an initial crack phenotype parameter;

[0133] S500: obtaining crack propagation simulation information through initial crack parameters, a comprehensive finite element model and material physical parameters, wherein the crack propagation simulation information at least includes crack propagation speed, maximum stress and maximum temperature;

[0134] S600: Based on the crack propagation simulation information and the equipment processing parameters, an optimization model is constructed and the equipment processing parameters are optimized to obtain the optimal processing parameters of the equipment;

[0135] S700: Inferring the processing parameters of the current period based on the crack risk prediction model to obtain the crack risk prediction value of the current period; based on the crack risk prediction value of the current period, the optimal processing parameters of the current period and the processing parameters of the current period, adjusting the actual optimal processing parameters of the current period to achieve control of the cracks of the ceramic bearing roller.

[0136] In S100, the physical parameters of the ceramic bearing roller material are obtained, and the physical parameters of the ceramic bearing roller material include elastic modulus , Poisson's ratio ,density , thermal expansion coefficient , thermal conductivity Specific heat capacity , fracture toughness .

[0137] Stress obtained by tensile test With strain , and then the elastic modulus is obtained and Poisson's ratio , which is expressed as follows:

[0138]

[0139] in, represent the lateral strain and longitudinal strain respectively.

[0140] Temperature changes obtained through thermal expansion experiments Length variation of ceramic bearing roller material under , and then the thermal expansion coefficient is obtained , which is expressed as follows:

[0141]

[0142] in, Indicates the initial length of the ceramic bearing roller material.

[0143] density Obtained by measuring mass and volume. Thermal conductivity is obtained by using the steady-state method through thermal property experiments. , using differential scanning calorimetry to obtain the specific heat capacity Through fracture tests, using single-edge notched beam specimens, the fracture toughness .

[0144] In S200, a two-dimensional microscopic image sequence of the ceramic bearing roller is obtained and preprocessed, and then a three-dimensional microstructure digital twin model is constructed. The three-dimensional microstructure digital twin model is subjected to finite element analysis and combined with the material physical parameters to obtain an initial finite element model of the ceramic bearing roller. This step aims to establish a digital twin model that can truly reflect the microstructure and physical properties of the ceramic material, and provide accurate data support for subsequent simulation analysis and crack prediction. The following steps are included:

[0145] S210: Scan the ceramic bearing roller sample layer by layer to obtain a two-dimensional microscopic image sequence of the ceramic bearing roller at different depths. Polish and etch the ceramic bearing roller sample to prepare a sample suitable for microscopic observation. Use a scanning electron microscope (SEM) and a projection electron microscope (TEM) to scan the sample layer by layer to obtain a two-dimensional microscopic image sequence of the ceramic bearing roller at different depths. ,in is the number of layers, the image sequence Used for subsequent ceramic roller image processing and 3D reconstruction.

[0146] S220: 2D microscopic image sequence Preprocessing is performed to obtain a preprocessed two-dimensional microscopic image sequence, the preprocessing includes, in sequence, using a Gaussian filter for denoising, performing a convolution operation on each image in the image sequence, using a histogram equalization to enhance the contrast of the image, segmenting the images in the two-dimensional microscopic image sequence based on a set threshold to achieve binarization of the two-dimensional microscopic image, using morphological operations (such as opening and closing operations) to remove isolated pixels and fill holes in the binary two-dimensional microscopic image, and the preprocessed two-dimensional microscopic image sequence is used for three-dimensional reconstruction.

[0147] S230: Based on the preprocessed two-dimensional microscopic image sequence and combined with the image-based three-dimensional reconstruction technology, a three-dimensional microstructure digital twin model of the ceramic bearing roller is obtained.

[0148] In this embodiment, the image-based three-dimensional reconstruction technology includes image registration and fusion technology (precisely aligning and deeply fusing two-dimensional microscopic images of different depths, and integrating the information of multiple image sequences into a three-dimensional model), three-dimensional reconstruction technologies such as stereo vision or structured light (the algorithm first restores the three-dimensional point cloud data of the object, and then constructs a three-dimensional microstructure model of the object after processing the point cloud data), artificial intelligence models such as convolutional neural networks (extracting features from two-dimensional microscopic images and predicting the three-dimensional structure of the object. By training a large amount of data, the model can accurately reconstruct the three-dimensional microstructure of the object) and Marching Cubes algorithm (isosurface extraction algorithm, generating a three-dimensional mesh model from volume data), etc. This application is not limited to this. The three-dimensional microstructure digital twin model includes the microstructure characteristics of the ceramic bearing roller, and the microstructure characteristics include grains, pores, and grain boundaries.

[0149] Compared with other image-based 3D reconstruction technologies mentioned above, the MarchingCubes algorithm has the advantages of good visualization, algorithm stability, and wide application fields. This embodiment is based on the preprocessed 2D microscopic image sequence and combined with the MarchingCubes algorithm to construct a 3D microstructure digital twin model of the ceramic bearing roller. Specifically, each image in the preprocessed 2D microscopic image sequence is taken as a layer, stacked according to the preset layer spacing, to form a 3D voxel matrix. , combined with the MarchingCubes algorithm, the voxel matrix Perform isosurface extraction to generate the corresponding 3D microstructure digital twin model .

[0150] S240: Perform finite element analysis on the three-dimensional microstructure digital twin model to obtain an initial finite element model of the ceramic bearing roller, wherein the finite element analysis includes at least meshing and material property assignment, different regions are formed through meshing, and corresponding material physical parameters are assigned to different regions.

[0151] Comprehensively consider factors such as the mesh unit type, mesh accuracy and computational efficiency, meshing algorithm and tools, and mesh quality inspection of finite element analysis, determine the mesh properties, mesh density, meshing strategy, and meshing software of finite element analysis, and perform meshing of finite element analysis on the three-dimensional microstructure digital twin model. The mesh unit type used in this embodiment is tetrahedral unit, and the Mesh module of ABAQUS is used for meshing. Refined meshes are used in areas with rich microstructural details (such as grain boundaries and around pores), and coarser meshes are used in areas with simple structures and small stress gradients.

[0152] The physical parameters of the ceramic material of the ceramic bearing rotor are assigned to the material properties of different regions in the meshed finite element model to obtain the initial finite element model of the ceramic bearing roller. In this embodiment, the elastic modulus of the ceramic material is given inside the grain , Poisson's ratio The grain boundaries are set as weakened areas and assigned different material parameters, and the pores are set as voids or assigned the material properties of air (the elastic modulus is close to zero).

[0153] Initial finite element model It is a finite element model of the three-dimensional microstructure digital twin model after fine mesh division and assignment of material properties, which is used for simulation calculations.

[0154] In S300, mechanical boundary conditions and thermal boundary conditions are applied to the initial finite element model to obtain a comprehensive finite element model, and the thermal-mechanical coupling simulation data is obtained by combining the physical parameters of the material, wherein the thermal-mechanical coupling simulation data at least includes a final temperature field, a final stress field, and a set of locations of high crack risk areas, including the following steps:

[0155] S310: Apply mechanical boundary conditions to the initial finite element model to achieve displacement constraints, and obtain a first finite element model. In this embodiment, the mechanical boundary conditions include displacement constraints, application of external forces and external stresses, and load application methods.

[0156] The displacement constraint is imposed to prevent the initial finite element model from rigid body displacement, that is, overall movement or rotation, during the simulation process, and some nodes of the model need to be fixed. The node displacement of the bottom or specific surface of the initial finite element model is set to 0, so that these nodes will not move during the simulation process, thereby ensuring the stability of the initial finite element model.

[0157] The application of external force is to apply concentrated force or distributed force from the outside to certain specific areas (such as the top or the load-bearing surface) of the initial finite element model. Distributed force is usually used to simulate uniformly acting forces such as gravity.

[0158] The application of external stress is to apply stress loads on the boundary surface of the model. This load is usually used to simulate boundary conditions such as pressure and tension.

[0159] Loading methods include static load and dynamic load. Static load means that the load applied to the initial finite element model is constant and does not change with time. It is usually used to simulate gravity, fixed support force, etc. in structural static analysis. Dynamic load means that the load applied to the initial finite element model changes with time. It can be periodic, impact or other forms of variable loads. It is usually used to simulate dynamic response problems such as vibration analysis, impact analysis, fatigue analysis, etc.

[0160] S320: Applying thermal boundary conditions to the first finite element model to obtain a comprehensive finite element model, wherein the thermal boundary conditions include temperature boundary conditions and heat flow boundary conditions.

[0161] The temperature boundary condition is used to specify the temperature setting in the thermal interaction between the first finite element model and the external environment, including setting the temperature on certain surfaces or regions of the first finite element model to a fixed temperature, setting the convection heat transfer coefficient and the ambient temperature to simulate the heat exchange between the surface of the first finite element model and the external environment. The convection heat transfer coefficient represents the efficiency of heat transfer to the surface of the first finite element model through the fluid boundary layer, and the convection heat transfer coefficient depends on factors such as the properties of the fluid, the flow rate, and the roughness of the model surface. The ambient temperature represents the temperature of the external environment of the first finite element model, which is usually known or can be obtained through experimental measurement.

[0162] Heat flux boundary conditions are used to describe the flow or injection of heat on the surface of the first finite element model, including setting heat flux density and internal heat source. Heat flux density means applying heat flux density on certain surfaces of the first finite element model, which represents the amount of heat passing through a unit area per unit time. Internal heat source means specifying a heat source item inside the material, which represents the heat generated per unit volume.

[0163] S330: Obtain thermal-mechanical coupling simulation data by integrating the finite element model and material physical parameters. The specific process includes:

[0164] S331: Constructing a control equation, which includes a mechanical equilibrium equation, a heat conduction equation, and a constitutive relationship equation. Based on the physical parameters of the material and the stress-strain (displacement) relationship of the material, a mechanical equilibrium equation is constructed, which is expressed as follows:

[0165]

[0166] in, represents stress, represents body density, Represents the density of ceramic materials, represents displacement, Indicates time.

[0167] Based on the relationship between material physical parameters and material heat conduction, the initial heat conduction equation is constructed as follows:

[0168]

[0169] in, represents the specific heat capacity of ceramic materials, represents temperature, which is a function of space and time, is the partial derivative of temperature with respect to time, indicating the rate of change of temperature, represents the thermal conductivity of ceramic materials, represents the heat source term, represents the thermal conductivity term, which indicates the process of heat conduction through the material. Thermal conduction can be classified into steady state or transient state. It is steady-state heat conduction.

[0170] The mechanical equilibrium equation and the heat conduction equation are related through the thermal strain constitutive relationship equation to realize thermal-mechanical coupling. The constitutive relationship equation is expressed as follows:

[0171]

[0172] in, represents the elastic stiffness matrix, represents the total strain, represents the thermal strain caused by temperature change, represents the matrix transpose, represents the thermal expansion coefficient of ceramic materials, represents the reference temperature, Represents the identity matrix.

[0173] S332: Combined with the comprehensive finite element model, the control equation is solved by finite element discretization and time discretization to obtain the final temperature field, the final stress field and the location set of high crack risk areas, including the following steps:

[0174] (1) Solve the control equations through the integrated finite element model to obtain the final displacement field and final stress field, including:

[0175] Setting the initial values ​​of the comprehensive finite element model, including setting the initial temperature field and initial displacement field;

[0176] By combining the finite element model and initial values, the heat conduction equation is solved to obtain a new temperature field;

[0177] By integrating the finite element model and the new temperature field, the constitutive relationship equation is solved to obtain a new stress field;

[0178] Through the comprehensive finite element model, the mechanical equilibrium equation is solved according to the new stress field to obtain the new displacement field and the new displacement field.

[0179] The above process is iterated until a convergence condition is met to obtain a final temperature field and a final stress field. The convergence condition includes that a relative error of a temperature field change and / or a stress field change is less than a preset threshold.

[0180] (2) Based on the final stress field, the location set of high crack risk areas is obtained, including:

[0181] Based on the final stress field, the maximum principal stress in each region is obtained, which is expressed as follows:

[0182]

[0183] in, represents the determinant, represents the final stress field, represents the eigenvalue of the final stress field, represents the 3*3 identity matrix, Represents the principal stress of the final stress field. The principal stress includes and ,in , so the maximum principal stress is .

[0184] The locations where the maximum principal stress in the comprehensive finite element model is not less than the stress threshold are identified, and the location set of high crack risk areas is obtained, which is expressed as follows:

[0185]

[0186] in, represents the maximum principal stress, represents the stress threshold, which is the failure stress (tensile strength) of the material. Indicates the location point, Represents the location set of high crack risk areas, including all locations where the maximum principal stress is greater than or equal to the stress threshold , that is, all the locations where the maximum principal stress in the comprehensive finite element model exceeds the material failure stress, The locations (elements or nodes) in the image are the high crack risk areas.

[0187] The thermal-mechanical coupling simulation data includes the final temperature field, the final stress field, and a set of locations of high crack risk areas. The thermal-mechanical coupling simulation data is used to train the next artificial intelligence model and guide crack extension simulation. When training the artificial intelligence model, the set of locations of high crack risk areas is used as label data to guide the model to learn crack initiation characteristics.

[0188] In 400, a crack initiation pre-training model is constructed and trained through thermal-mechanical coupling simulation data to obtain a crack initiation model, and then obtain initial crack parameters, wherein the initial crack parameters at least include an initial crack position, an initial crack direction, and an initial crack phenotype parameter, including the following steps:

[0189] S410: Construct a crack initiation pre-training model based on a three-dimensional space artificial intelligence model. Artificial intelligence models suitable for processing three-dimensional space data include 3D convolutional neural networks (3D CNN), graph neural networks (GNN), physics-based deep learning models, variational autoencoders (VAE), generative adversarial networks (GAN), neural networks based on attention mechanisms, and hybrid models, which are not limited in this application.

[0190] In this embodiment, a 3D CNN model is used to construct a crack initiation pre-training model, including: stacking multiple convolutional layers to extract spatial features, using ReLU as the activation function, flattening the convolutional layer output and connecting it to the fully connected layer, using the Sigmoid activation function in the output layer, and mapping the output to interval, indicating the crack risk probability, and the convolution kernel size is set to The number of convolutional layers is determined according to the model complexity and the amount of data. In this embodiment, 3 is selected. The loss function adopts the binary cross entropy loss function and uses the Adam optimizer.

[0191] S420: Training and testing the crack initiation pre-training model using thermal-mechanical coupling simulation data to obtain a crack initiation model, including the following steps:

[0192] (1) Based on the preset data division ratio, the thermal-mechanical coupling simulation data is divided into a training set, a validation set, and a test set, including:

[0193] Based on the final stress field And the final temperature field , construct a multi-channel three-dimensional input tensor, which is the input for the crack initiation pre-training model training. The final stress field Contains the normal stress component of the stress field and the shear stress component of the stress field , then the initial multi-channel three-dimensional input tensor , which is expressed as follows:

[0194]

[0195] If the space grid is ,but The tensor dimension is ,in (Number of channels). Preprocess the initial multi-channel three-dimensional input tensor to obtain a multi-channel three-dimensional input tensor The preprocessing includes normalizing each channel of the initial multi-channel three-dimensional input tensor, dividing the large-sized three-dimensional data into small-sized sub-blocks to improve the training efficiency and generalization ability of the model, and balancing the positive and negative samples to reduce the model training bias.

[0196] A collection of locations of high crack risk areas based on thermal-mechanical coupling simulation data , construct the crack risk label tensor, which is used as the output of artificial intelligence model training. , construct the corresponding crack risk label tensor , crack risk label tensor The dimension is , which is expressed as follows:

[0197]

[0198] The multi-channel three-dimensional input tensor and crack risk label tensor are divided into training set, validation set and test set based on the preset division ratio. The commonly used division ratio is training set , validation set , test set .

[0199] (2) The crack initiation pre-training model is trained and tested through the training set, validation set and test set to obtain a trained crack initiation model, including:

[0200] The initial crack initiation pre-trained model is set as the best model;

[0201] Divide the training set into batches to calculate the model prediction output;

[0202] Based on the predicted output of the model and the loss function, we get the loss of the current batch.

[0203] Based on the loss of the current batch, calculate the gradient of the loss to the model parameters to achieve back propagation;

[0204] Update the model's parameters using an optimization algorithm;

[0205] Calculate the loss and evaluation indicators on the validation set. The evaluation indicators include at least accuracy and AUC.

[0206] If the performance on the validation set is better than the previous best model, then the best model is updated.

[0207] According to the evaluation indicators of the model, adjust the model's hyperparameters (such as learning rate, batch size, number of network layers) and retrain the model until the model performance reaches the preset conditions.

[0208] The crack initiation model can accurately predict the location set of high crack risk areas based on stress field and temperature field.

[0209] S430: Input the thermal-mechanical coupling simulation data to be tested into the crack initiation model to obtain a predicted set of locations of high crack risk areas. Obtain new stress field distribution and new temperature field distribution, construct a multi-channel three-dimensional input tensor and pre-process the multi-channel three-dimensional input tensor, input the pre-processed multi-channel three-dimensional input tensor into the crack initiation model, and obtain a set of locations of high crack risk areas predicted by the model.

[0210] S440: Based on the predicted high crack risk area location set, initial crack parameters are obtained, wherein the initial crack parameters at least include initial crack location, initial crack direction, and initial crack phenotype parameters, which are expressed as follows:

[0211] (1) Initial crack position: Calculate the centroid of the predicted high crack risk area position set to obtain the initial crack position. The initial crack position is the initial crack position, which is expressed as follows:

[0212]

[0213] in, The centroid of the predicted high crack risk area location set is the initial crack location. represents the predicted high crack risk area location set, The number of voxels representing the set of locations of predicted high crack risk areas.

[0214] (2) The initial crack phenotype parameters include the initial crack shape and initial crack size. In this embodiment, according to the predicted high crack risk area location set The spatial distribution of the initial crack shape is fitted, and then the size of the initial crack (the size of the fitted shape) is obtained. For example, according to the predicted high crack risk area location set, the crack shape is fitted into an ellipsoid, and then the size information such as the major axis length and minor axis length is obtained.

[0215] (3) Initial crack direction. In this embodiment, the predicted high crack risk area location set is calculated as follows: The maximum principal stress is obtained, and then the average value of the maximum principal stress direction is obtained, and the average value of the maximum principal stress direction is normalized to a unit vector, and the unit vector is the initial crack direction, comprising the following steps:

[0216] The maximum principal stress of the high crack risk area location set is obtained by using the above method to solve the maximum principal stress of the high crack risk area location set. , I will not go into details here.

[0217] Based on the maximum principal stress , the average value of the maximum principal stress direction in the predicted high crack risk area location set area is obtained, which is expressed as follows:

[0218]

[0219] in, represents the average value of the maximum principal stress direction of the predicted high crack risk area location set, It is the eigenvector corresponding to the maximum principal stress of the high crack risk area location set.

[0220] The average value of the direction of the maximum principal stress Normalized to unit vector , the unit vector is the initial crack direction.

[0221] The extracted initial crack parameters can provide accurate initial conditions for crack propagation simulation, so that the crack propagation simulation can accurately simulate the crack initiation and propagation process.

[0222] In S500, crack propagation simulation information is obtained through initial crack parameters, a comprehensive finite element model and material physical parameters, wherein the crack propagation simulation information at least includes crack propagation speed, maximum stress and maximum temperature, including the following steps:

[0223] S510: Using the initial crack position as a reference system, matching the initial crack phenotype parameters to the comprehensive finite element model, and then matching the crack surface normal vector of the comprehensive finite element model to the initial crack direction and refining the mesh of the crack area to update the comprehensive finite element model, and defining the crack propagation condition, crack propagation direction and crack propagation increment to obtain the crack propagation finite element model, specifically including:

[0224] (1) Matching the initial crack parameters to the integrated finite element model and refining the mesh of the initial crack area to update the integrated finite element model, including:

[0225] The initial crack position Match to the corresponding position of the comprehensive finite element model.

[0226] Taking the initial crack position in the comprehensive finite element model as the reference system, the initial crack shape and size are matched to the comprehensive finite element model. In this embodiment, the initial crack is defined as a semi-elliptical crack, which is expressed as follows:

[0227]

[0228] in, represents the depth of the initial crack, Represents the surface half-length of the initial crack.

[0229] Crack direction: the crack surface normal vector of the initial crack matched into the comprehensive finite element model is matched as the initial crack direction, that is, the crack extends along the initial crack direction. The direction of the crack surface normal vector is orthogonal to the eigenvector of the maximum principal stress in the predicted high crack risk area position set. The magnitude of the crack surface normal vector is 1 and only represents the direction.

[0230] The mesh of the integrated finite element model is updated to increase the mesh density near the initial crack tip region to ensure computational accuracy in order to capture stress singularities.

[0231] (2) Define the crack propagation conditions, crack propagation direction, crack propagation increment and crack location, and obtain the crack propagation finite element model, which includes:

[0232] The crack extension condition includes that crack extension is performed if the current crack energy release rate is not less than the energy release threshold, and the energy release rate is expressed as follows:

[0233]

[0234] in, represents the current energy release rate, represents the equivalent elastic modulus of the ceramic bearing roller, represents the stress intensity factor of the opening mode of the previous iteration, represents the stress intensity factor of the slip mode of the previous iteration, represents the stress intensity factor of the tearing mode of the previous iteration, where , and , respectively expressed as follows:

[0235]

[0236]

[0237]

[0238] in, represents the maximum principal stress in the normal direction of the current crack tip, represents the crack length of the previous iteration, represents the tangential shear stress of the stress tensor at the current crack tip, The normal shear stress represents the stress tensor at the current crack tip. The initial stress intensity factor , and According to the initial crack length get.

[0239] Among them, the energy release threshold is the critical energy release rate of the ceramic material, which is expressed as follows:

[0240]

[0241] in, represents the critical energy release rate of ceramic materials, is the fracture toughness of the ceramic material, is the equivalent elastic modulus.

[0242] The crack propagation direction is expressed as follows:

[0243]

[0244]

[0245] in, represents the current crack propagation direction, represents the crack growth direction of the previous iteration, represents the crack growth angle of the previous iteration.

[0246] The crack growth rate and crack growth increment are expressed as follows:

[0247]

[0248]

[0249] in, represents the current crack growth rate, represents the crack growth rate constant, It represents the critical energy release rate of the material of the ceramic bearing roller, that is, the energy release threshold. represents the crack growth index, represents the current crack growth increment, Indicates the time difference.

[0250] The crack location is shown below:

[0251]

[0252] in, is the current crack position, is the crack position of the previous iteration, is the unit vector of the crack propagation direction.

[0253] The above-mentioned methods for updating the integrated finite element model include the extended finite element method (XFEM), virtual crack closure technology (VCCT), interaction integration method (such as Integration method), etc., this application does not limit it.

[0254] S520: Based on the crack propagation finite element model, the control equation is iteratively solved through space discretization and time discretization. Based on the data of each iteration in the solution process, the crack propagation simulation information is obtained. The solution process is the crack propagation simulation process, which specifically includes:

[0255] (1) When the crack propagation conditions are met, crack propagation is carried out to obtain the current crack propagation direction, crack propagation speed, crack propagation increment, crack position, stress field, and temperature field, and then update the crack propagation finite element model.

[0256] (2) Based on the updated crack propagation finite element model, the control equations are re-solved until the solution termination condition is met. The solution termination condition can be based on whether the crack propagation length reaches a preset crack propagation length, whether the simulation time reaches a preset simulation time, or whether the crack position reaches a preset position.

[0257] (3) Based on the stress field and temperature field of each iteration, the maximum stress and maximum temperature are obtained.

[0258] Crack growth simulation information also includes crack growth path , stress field evolution , temperature field evolution .

[0259] Crack propagation path is the crack position of each iteration, that is, the set of crack positions of each iteration. Based on the crack propagation path, a crack propagation trajectory can be drawn to intuitively observe the crack propagation path.

[0260] Stress field evolution , temperature field evolution is the change of stress field and temperature field with time, characterizes the evolution of stress field and temperature field, and is the stress field of each iteration in the iterative process. , Temperature field Composed of a collection.

[0261] Furthermore, in order to ensure the accuracy and reliability of the simulation results, the credibility of the model can be improved through model verification and validation, including but not limited to using different grid densities for simulation to ensure the convergence of the results, and checking the impact of the time period length on the simulation results by adjusting the time period length to ensure that the time discretization error is within an acceptable range.

[0262] Based on the initial crack parameters and the comprehensive finite element model, a thermal-mechanical coupled crack propagation simulation is performed to obtain crack propagation simulation information, which provides key data for processing parameter optimization.

[0263] In S600, based on the crack propagation simulation information and the equipment processing parameters, an optimization model is constructed and the equipment processing parameters are optimized to obtain the optimal processing parameters of the equipment, including the following steps:

[0264] S610: Construct an optimization model based on the crack propagation simulation information and equipment processing parameters.

[0265] (1) Taking the processing parameters of the equipment as the optimization variables, the optimization variable model is expressed as follows

[0266]

[0267] in, represents the optimization variable, i.e., the equipment processing parameters, Indicates the cutting speed, Indicates the feed rate, Indicates the cutting depth, Indicates the coolant flow rate.

[0268] (2) Comprehensively considering the risk of cracks and the processing cost value, an optimization target model is established. In this embodiment, a weighted sum method is used to comprehensively combine the crack risk and the processing cost value to form an optimization target model. The processing cost value includes processing time, tool wear and energy consumption. The optimization target model is expressed as follows:

[0269]

[0270]

[0271]

[0272] in, represents the optimization goal, represents the weight coefficient, and , represents the first crack risk value, Indicates the processing cost value. represents the maximum stress, is the crack growth rate, Indicates the maximum temperature, They all represent weight coefficients, reflecting the importance of each factor. They represent the processing time cost, tool wear cost, and energy consumption cost respectively. Represent unit time cost, unit wear cost, and unit energy cost respectively. Indicates the processing time, Indicates the feed rate, Indicates the cutting speed, Indicates the length of the ceramic bearing roller, Indicates the wear of the machining tool, Indicates energy consumption.

[0273] (3) The constraint model includes the processing parameter constraint model and the processing quality constraint model, which are expressed as follows:

[0274]

[0275]

[0276] in, Indicates the surface roughness of ceramic bearing rollers. Indicates the machining dimensional accuracy required for ceramic bearing rollers. Indicates the wear of the machining tool, Indicates the maximum and minimum values ​​of the corresponding parameters.

[0277] S620: Based on the optimization model, combined with the first optimization (genetic algorithm) algorithm, a preliminary solution of the optimization model (preliminary optimization of processing parameters) is obtained.

[0278] The main purpose of the first optimization algorithm is to quickly find a feasible preliminary solution and provide a starting point for subsequent optimization. Therefore, the first optimization algorithm should have the characteristics of speed (being able to find a solution in a short time, even if the solution is not globally optimal), simplicity (the algorithm is relatively simple to implement and does not require too many computing resources) and robustness (being able to stably find a solution under different initial conditions). The algorithms that can be used for the first optimization algorithm include grid search, random search and heuristic algorithms. Grid search can systematically search the parameter space. When the parameter range is small and continuous, it can be used as a simple and fast preliminary solution search method; random search algorithms can randomly select parameter combinations and find a feasible solution through multiple attempts. It is simple and easy to implement; heuristic algorithms, such as simplified genetic algorithms, can use heuristic rules to guide the search, which can balance the search speed and the quality of the solution to a certain extent. This application does not limit the specific first optimization algorithm used.

[0279] In this embodiment, a genetic algorithm (GA) is selected as an example of the first optimization algorithm, aiming to demonstrate and optimize the process of preliminary solution to the optimization problem through the genetic algorithm. The first optimization algorithm obtains a better set of processing parameters and provides a preliminary solution for subsequent fine optimization, including the following steps:

[0280] (1) Initialize the population. Randomly generate an initial population, that is, an initial combination of equipment processing parameters. The population size can be determined based on the complexity of the problem and computing resources.

[0281] (2) The encoding of the population is controlled by the constraint model, and the population, i.e., the equipment processing parameters, are encoded to form chromosomes or individuals in the genetic algorithm. The encoding method can be selected according to the characteristics of the specific problem, such as binary encoding, real number encoding, symbol encoding, etc.

[0282] (3) The optimization target value is used as the fitness to evaluate the quality of each individual and determine whether the optimization criteria are met (for example, the maximum fitness of individuals in the population exceeds the preset value). If so, the best individual and the optimal solution it represents are output and the algorithm ends; if not, proceed to the next step.

[0283] (4) According to fitness, select a part of individuals as parents to generate the next generation. Individuals with high fitness have a high probability of being selected. On the contrary, individuals with low fitness have a low probability of being selected and may even be eliminated. The selection operation can adopt methods such as roulette selection and tournament selection.

[0284] (5) Perform a crossover operation on the selected parent individuals to generate new excellent offspring individuals. The crossover operation can be single-point crossover, multi-point crossover, consistent crossover, sequential crossover, and periodic crossover, among which single-point crossover is the most widely used.

[0285] (6) Perform mutation operations on the newly generated offspring individuals to increase the diversity of the population. The mutation operation can be carried out by random mutation, non-uniform mutation and other methods.

[0286] (7) Calculate the fitness of each individual in the new population and sort them according to fitness.

[0287] (8) Determine whether the termination condition is met. If so, output the optimal solution; otherwise, return to step (4) to continue iterating. The termination condition includes one or more of the following: reaching the maximum number of iterations, the maximum fitness of individuals in the population exceeding the set value, etc. The optimal solution obtained by the genetic optimization algorithm is the preliminary solution of the optimization model (preliminary equipment processing parameters).

[0288] S630: Based on the optimization model and the preliminary solution, combined with the second optimization algorithm (particle swarm algorithm), a global optimal solution of the optimization model is obtained, and the global optimal solution is the optimal processing parameter of the equipment;

[0289] The second optimization algorithm needs to further search on the basis of the preliminary solution to find the global optimal solution or a solution close to the global optimal solution. Therefore, the second algorithm should have the characteristics of accuracy (being able to find a high-quality global optimal solution), convergence (with the increase of the number of iterations, the quality of the solution should gradually improve and eventually converge to a stable solution) and adaptability (being able to handle complex constraints and objective functions). The algorithms that can be used for the second optimization algorithm include particle swarm, simulated annealing, gradient descent, etc. The particle swarm algorithm can simulate the group behavior of bird flocks or fish schools to search for the optimal solution and has a faster convergence speed; the simulated annealing algorithm is an optimization algorithm based on the physical annealing process, which is suitable for dealing with problems with multiple local optimal solutions; the gradient descent method or its variants, for differentiable objective functions, can gradually approach the optimal solution by calculating the gradient. This application does not limit the specific second optimization algorithm used.

[0290] In this embodiment, a particle swarm algorithm (PSO) is selected as an example of the second optimization algorithm, and the purpose is to specifically demonstrate and optimize the preliminary solution through the particle swarm algorithm to find the global optimal combination of processing parameters, including the following steps:

[0291] (1) Randomly generate an initial particle swarm, including the initial position and velocity of the particle swarm, and then obtain the initial individual optimal position of the particle and the initial global optimal position of the particle swarm. Each particle position represents a device processing parameter. , the particle velocity indicates the direction and rate of particle position update, the initial individual optimal position of a particle is its initial position, and the initial global optimal position of a particle swarm is the position of a random particle in the initial particle swarm.

[0292] (2) Obtaining the optimization target value corresponding to the current position (equipment processing parameter) of each particle by optimizing the target model, wherein the optimization target value is the fitness of the current position of the particle;

[0293] (3) According to the fitness, if the fitness of the current position is better than the fitness of the individual optimal position, then the individual optimal position (the position of each particle with the minimum fitness in its history) is updated; if the fitness of any particle's current position is better than the fitness of the global optimal position, then the global optimal position (the position of the particle with the smallest fitness among all particles so far) is updated;

[0294] (4) According to the particle's current position, speed, individual optimal position, and global optimal position, update the particle's speed and position to obtain an updated particle speed and an updated particle position;

[0295] (5) Check whether the particle position satisfies the constraint model. If not, make corrections and adjustments.

[0296] (6) Iteration. If the termination condition is met, the iteration ends and the final global optimal position is output. The final global optimal position is the optimal processing parameter. , if the termination condition is not met, repeat the above steps (2) to (5) until the termination condition is met. The termination condition can be reaching the maximum number of iterations, or the fitness change is less than a certain threshold, or other forms.

[0297] In this embodiment, when selecting the first optimization algorithm and the second optimization algorithm, it is also necessary to consider that the two algorithms should have certain complementarity: complementarity between global search and local search, the first optimization algorithm should have global search capabilities to find a relatively good preliminary solution, and the second optimization algorithm should have strong local search capabilities to find the optimal solution near the preliminary solution; complementarity of algorithm characteristics, for example, genetic algorithms are good at global search but may converge slowly, while the gradient descent method converges quickly but is prone to fall into local optimality. Therefore, the genetic algorithm can be used as the first optimization algorithm and the gradient descent method can be used as the second optimization algorithm; complementarity of computational complexity, the first optimization algorithm can select an algorithm with low computational complexity and easy implementation to quickly obtain a preliminary solution, while the second optimization algorithm can select an algorithm with high computational complexity but better effect to perform fine optimization.

[0298] The crack propagation simulation provides key physical quantities, such as maximum stress, crack propagation rate and maximum temperature, which are related to the processing parameters. By associating the physical quantities with the processing parameters, the optimization objective function is established, and the optimization objectives and constraints are clarified. The first optimization algorithm is used to preliminarily search the global solution space and provide a better solution. The second optimization algorithm performs a fine search based on the first optimization algorithm. The combination of the first optimization algorithm and the second optimization algorithm improves the accuracy and quality of the solution.

[0299] In S700, the processing parameters of the current period are inferred based on the crack risk prediction model to obtain the crack risk prediction value of the current period, and based on the crack risk prediction value of the current period, the optimal processing parameters of the current period and the processing parameters of the current period, the actual optimal processing parameters of the current period are adjusted to achieve the control of the cracks of the ceramic bearing roller, including the following steps:

[0300] S710: Build a pre-trained model for crack risk prediction based on the time series artificial intelligence model.

[0301] In this embodiment, the artificial intelligence models suitable for processing time series data include machine learning models (such as Prophet, XGBoost, gradient boosting machine GBM and random forest RF, etc.) and deep learning models (such as LSTM, Transformer, Seq2Seq, TCN, DeepAR, etc.). The Prophet model has a good effect when processing time series data with periodicity, trend and holiday effects. The XGBoost model has high accuracy when processing nonlinear relationships and regression problems of time series data. Gradient boosting machine (GBM) and random forest (RF) are suitable for processing nonlinear relationships and complex time series data. The LSTM model is specifically used to process time series data, can capture long-term dependencies, and performs well in time series prediction tasks. The Transformer model can process long-series time data and improve prediction accuracy by capturing global dependencies in time series. The Seq2Seq model is suitable for processing input and output time series of different lengths, and can be predicted by learning the mapping relationship between input sequences and output sequences. TCN has strong time series modeling capabilities and processes time series data through operations such as one-dimensional convolution and causal convolution. DeepAR is suitable for processing time series data with uncertainty. This application does not limit the specific time series artificial intelligence model.

[0302] In this embodiment, the Transformer model is selected as the time series artificial intelligence model. In terms of the model architecture, the feature vector of each time step is converted into an embedding vector, position encoding is added to retain the order information of the time step in the sequence, a multi-head self-attention mechanism is adopted, a feedforward network is applied independently to each position, residual connections and layer normalization are added after each sub-layer to improve the stability and training efficiency of the model, multi-layer encoders are stacked, and there are at least 6 layers of multi-head self-attention and feedforward networks. Through the fully connected layer and the Sigmoid activation function, the output of the Transformer is mapped to the crack risk probability. In terms of model training, the binary cross entropy loss function for binary classification tasks is used as the loss function of the model, and the Adam optimizer is used to optimize the model parameters.

[0303] S720: Training a crack risk prediction pre-training model based on the historical first processing parameter of the equipment to obtain a crack risk prediction model, including the following steps:

[0304] (1) Obtain the historical first processing parameters of the equipment. The historical first processing parameters of the equipment are the data of each period during the processing. The first processing parameters include cutting force data, vibration data, temperature data and other sensor data (such as acoustic emission, cutting fluid status, etc.) and crack risk value. Among them, cutting force data, vibration data, temperature data and other sensor data are all obtained through sensors. Cutting force data includes main cutting force, feed force and radial force. Cutting force is the force generated by the interaction between the tool and the workpiece during the processing process. It can reflect whether abnormal load or uneven cutting occurs during the processing process, which will cause tool damage or stress concentration, thereby increasing the risk of cracks. Vibration data includes mechanical vibration during the processing process, especially the contact between the tool and the workpiece, which will cause stability problems during the processing process, such as poor contact between the tool and the workpiece, machine tool resonance or other mechanical problems. These abnormal vibrations are early signs of crack formation. Temperature data reflects the heat generation during the processing process, including cutting zone temperature, tool temperature, workpiece temperature and coolant temperature. Excessive temperature will lead to increased tool wear, thermal deformation of the workpiece or cracks. By monitoring temperature data, potential overheating risks can be identified and corresponding control measures can be taken.

[0305] Crack risk value , which is expressed as follows:

[0306] .

[0307] (2) Cleaning and preprocessing (normalization or standardization) the historical first processing parameters to ensure that each feature is within a similar numerical range, dividing the cleaned and preprocessed historical first processing parameters into sequences of fixed length, and determining the training set, validation set, and test set;

[0308] (3) The crack risk prediction pre-training model is trained and verified using a training set, a validation set, and a test set. The trained crack risk prediction pre-training model is the crack risk prediction model.

[0309] The crack risk prediction model can predict the crack risk prediction value based on parameters in the processing process such as force, vibration, and temperature.

[0310] S730: Obtain a first processing parameter of the equipment in the current period, pre-process the first processing parameter in the current period, and input the parameter into a crack risk prediction model to obtain a crack risk prediction value in the current period.

[0311] S740: Based on the crack risk prediction value of the current period, the optimal processing parameters of the current period, and the processing parameters of the current period, the adjusted processing parameters of the current period are obtained, including the following steps:

[0312] (1) Based on the crack risk prediction value of the current period and the preset target crack risk level, the risk error is obtained, which is expressed as follows:

[0313]

[0314] in, represents the risk error, Indicates the preset target crack risk level, which is usually set to a low risk, such as equal to 0 or close to , Indicates the crack risk prediction value for the current period.

[0315] (2) Based on the risk error, the optimal processing parameters of the current period and the processing parameters of the current period, the theoretical processing parameter adjustment amount of the current period is obtained, which is expressed as follows:

[0316]

[0317] in, Indicates the theoretical processing parameter adjustment amount in the current period. represents the control gain, which determines the influence of risk error on the adjustment of processing parameters. Represents the weight coefficient, which is used to balance the impact of optimizing processing parameters and real-time adjustments. Indicates the optimal processing parameters for the current period, Indicates the processing parameters of the current period.

[0318] (3) Add the theoretical processing parameter adjustment amount and the current period processing parameters to obtain the theoretical optimal processing parameters for the current period.

[0319] (4) Limit the theoretical optimal processing parameters of the current period to the preset range, and obtain the adjusted processing parameters of the current period, which are expressed as follows:

[0320]

[0321] in, Indicates the processing parameters adjusted in the current period. Indicates the theoretical optimal processing parameters for the current period, and Represents the preset minimum and maximum values. The clip function limits the theoretical optimal processing parameters of the current period to between the preset minimum and maximum values.

[0322] S750: Adjust the actual optimal machining parameters of the current period based on the adjusted machining parameters of the current period, so as to achieve real-time dynamic adjustment of the machining parameters, thereby effectively reducing the crack risk of the ceramic bearing rollers and improving the machining quality and production efficiency.

[0323] Furthermore, new sensor data and corresponding crack risk values ​​can be collected from the actual processing process, and the crack risk prediction pre-trained model can be retrained to obtain a crack risk prediction model that is more in line with the current equipment processing conditions. The performance of the updated model on new data can be verified to ensure its generalization ability, and the verified model can be deployed to the real-time prediction system.

[0324] Example 2

[0325] An artificial intelligence-based ceramic bearing roller crack control system, such as Figure 2 As shown, it includes a material parameter acquisition module 100, a finite element model acquisition module 200, a thermal data acquisition module 300, a crack parameter acquisition module 400, a crack extension information acquisition module 500, an optimal processing parameter acquisition module 600 and an actual optimal processing parameter acquisition module 700.

[0326] The material parameter acquisition module 100 is used to acquire the material physical parameters of the ceramic bearing roller.

[0327] The finite element model acquisition module 200 is used to acquire a two-dimensional microscopic image sequence of the ceramic bearing roller and perform preprocessing, thereby constructing a three-dimensional microstructure digital twin model, performing finite element analysis on the three-dimensional microstructure digital twin model and combining it with material physical parameters to obtain an initial finite element model of the ceramic bearing roller.

[0328] The thermal data acquisition module 300 is used to apply mechanical boundary conditions and thermal boundary conditions to the initial finite element model to obtain a comprehensive finite element model, and to obtain thermal-mechanical coupling simulation data in combination with material physical parameters, wherein the thermal-mechanical coupling simulation data at least includes a final temperature field, a final stress field, and a set of locations of high crack risk areas.

[0329] The crack parameter acquisition module 400 is used to construct a crack initiation pre-training model and train it through thermal-mechanical coupling simulation data to obtain a crack initiation model, and then obtain initial crack parameters, wherein the initial crack parameters at least include initial crack position, initial crack direction and initial crack phenotype parameters.

[0330] The crack extension information acquisition module 500 is used to obtain crack extension simulation information through initial crack parameters, comprehensive finite element model and material physical parameters, wherein the crack extension simulation information at least includes crack extension speed, maximum stress and maximum temperature.

[0331] The optimal processing parameter acquisition module 600 is used to construct an optimization model and optimize the equipment processing parameters based on the crack propagation simulation information and the equipment processing parameters to obtain the optimal processing parameters of the equipment.

[0332] The actual optimal processing parameter acquisition module 700 is used to infer the processing parameters of the current period based on the crack risk prediction model to obtain the crack risk prediction value of the current period, and adjust the actual optimal processing parameters of the current period based on the crack risk prediction value of the current period, the optimal processing parameters of the current period and the processing parameters of the current period to achieve control of the cracks of the ceramic bearing rollers.

[0333] Various changes and modifications can be made without departing from the spirit and scope of the present invention, and all equivalent technical solutions also belong to the scope of the present invention.

[0334] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the various embodiments can be referenced to each other.

[0335] It will be appreciated by those skilled in the art that embodiments of the present invention may be provided as methods, devices, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented 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.

[0336] The present invention is described with reference to the flowcharts and / or block diagrams of the method, terminal device (system), and computer program product according to the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of the processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing terminal device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing terminal device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0337] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing terminal device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce a manufactured product including an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0338] These computer program instructions can also be loaded onto a computer or other programmable data processing terminal device so that a series of operating steps are executed on the computer or other programmable terminal device to produce a computer-implemented process, thereby providing instructions for executing on the computer or other programmable terminal device to implement the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0339] It should be noted that:

[0340] The "one embodiment" or "embodiment" mentioned in the specification means that the specific features, structures or characteristics described in conjunction with the embodiment are included in at least one embodiment of the present invention. Therefore, the phrases "one embodiment" or "embodiment" appearing in various places throughout the specification do not necessarily refer to the same embodiment.

[0341] In addition, it should be noted that the shapes and names of the parts and components of the specific embodiments described in this specification may be different. Any equivalent or simple changes made based on the structure, features and principles described in the patent concept of the present invention are included in the protection scope of the patent of the present invention. The technicians in the technical field of the present invention can make various modifications or supplements to the specific embodiments described or replace them in a similar manner, as long as they do not deviate from the structure of the present invention or exceed the scope defined by the claims, they should all fall within the protection scope of the present invention.

Claims

1. A method for controlling cracks in ceramic bearing rollers based on artificial intelligence, characterized in that: The following steps are involved: Obtain the material physical parameters of ceramic bearing rollers; Obtain a 2D microscopic image sequence of the ceramic bearing roller and perform preprocessing to build a 3D microstructure digital twin model. Perform finite element analysis on the 3D microstructure digital twin model and combine it with the material physical parameters to obtain the initial finite element model of the ceramic bearing roller. Apply mechanical boundary conditions and thermal boundary conditions to the initial finite element model to obtain a comprehensive finite element model, and combine the physical parameters of the material to obtain thermal-mechanical coupling simulation data, wherein the thermal-mechanical coupling simulation data at least includes a final temperature field, a final stress field, and a set of locations of high crack risk areas; A crack initiation pre-training model is constructed and trained through thermal-mechanical coupling simulation data to obtain a crack initiation model, and then obtain initial crack parameters, wherein the initial crack parameters at least include an initial crack position, an initial crack direction, and an initial crack phenotype parameter; Obtain crack propagation simulation information through initial crack parameters, comprehensive finite element model and material physical parameters, wherein the crack propagation simulation information at least includes crack propagation speed, maximum stress and maximum temperature; Based on the crack propagation simulation information and equipment processing parameters, an optimization model is constructed and the equipment processing parameters are optimized to obtain the optimal processing parameters of the equipment; Based on the crack risk prediction model, the processing parameters of the current period are inferred to obtain the crack risk prediction value of the current period, and based on the crack risk prediction value of the current period, the optimal processing parameters of the current period and the processing parameters of the current period, the actual optimal processing parameters of the current period are adjusted to achieve the control of the cracks of the ceramic bearing rollers; The step of applying mechanical boundary conditions and thermal boundary conditions to the initial finite element model to obtain a comprehensive finite element model and combining the physical parameters of the material to obtain thermal-mechanical coupling simulation data comprises the following steps: Applying mechanical boundary conditions to the initial finite element model to achieve displacement constraints, thereby obtaining a first finite element model; Applying thermal boundary conditions to the first finite element model to obtain a comprehensive finite element model, wherein the thermal boundary conditions include temperature boundary conditions and heat flow boundary conditions; Based on the comprehensive finite element model and material physical parameters, the thermal-mechanical coupling simulation data is obtained. The specific process includes: Constructing control equations, wherein the control equations include at least a mechanical equilibrium equation, a heat conduction equation, and a constitutive relationship equation, and solving the control equations by finite element discretization and time discretization in combination with a comprehensive finite element model to obtain a final temperature field, a final stress field, and a set of locations of high crack risk areas; The mechanical equilibrium equation, heat conduction equation and constitutive relationship equation are expressed as follows: in, represents stress, represents body density, represents the density of ceramic materials, represents displacement, Indicates time, represents the specific heat capacity of ceramic materials, Indicates temperature, represents the thermal conductivity of ceramic materials, represents the heat source term, represents the thermal conductivity term, represents the elastic stiffness matrix, represents the total strain, represents the thermal strain caused by temperature change, represents the matrix transpose, represents the thermal expansion coefficient of ceramic materials, represents the reference temperature, Represents the identity matrix.

2. The method for controlling cracks in ceramic bearing rollers based on artificial intelligence according to claim 1 is characterized in that: The method of obtaining a two-dimensional microscopic image sequence of a ceramic bearing roller and performing preprocessing to thereby construct a three-dimensional microstructure digital twin model, performing finite element analysis on the three-dimensional microstructure digital twin model and combining the physical parameters of the material to obtain an initial finite element model of the ceramic bearing roller comprises the following steps: The ceramic bearing roller is scanned layer by layer to obtain a sequence of two-dimensional microscopic images of the ceramic bearing roller at different depths; Preprocessing the two-dimensional microscopic image sequence to obtain a preprocessed two-dimensional microscopic image sequence, wherein the preprocessing includes one or more of denoising, convolution operation, equalization, segmentation, removing isolated pixels and filling holes; Based on the preprocessed two-dimensional microscopic image sequence and combined with three-dimensional reconstruction technology, the corresponding three-dimensional microstructure digital twin model is obtained; Finite element analysis is performed on the three-dimensional microstructure digital twin model to obtain an initial finite element model of the ceramic bearing roller, wherein the finite element analysis at least includes meshing and material property assignment, different regions are formed through meshing, and corresponding material physical parameters are assigned to different regions.

3. The method for controlling cracks in ceramic bearing rollers based on artificial intelligence according to claim 1, characterized in that: Obtaining the high crack risk area location set comprises the following steps: Based on the final stress field, the maximum principal stress in each region is obtained; The areas where the maximum principal stress in the comprehensive finite element model is not less than the stress threshold are identified, and the location set of high crack risk areas is obtained, which is expressed as follows: in, represents the location set of high crack risk areas, represents the maximum principal stress, represents the stress threshold, Indicates a location point.

4. The method for controlling cracks in ceramic bearing rollers based on artificial intelligence according to claim 1, characterized in that: The method of constructing a crack initiation pre-training model and training it through thermal-mechanical coupling simulation data to obtain a crack initiation model and then obtain initial crack parameters includes the following steps: Construct a crack initiation pre-training model based on a three-dimensional space artificial intelligence model; The crack initiation pre-training model is trained and tested through thermal-mechanical coupling simulation data to obtain a crack initiation model; Input the thermal-mechanical coupling simulation data to be tested into the crack initiation model to obtain a set of predicted high crack risk area locations; Based on the predicted high crack risk area location set, initial crack parameters are obtained, wherein the initial crack parameters at least include initial crack location, initial crack direction, and initial crack phenotype parameters.

5. The method for controlling cracks in ceramic bearing rollers based on artificial intelligence according to claim 4 is characterized in that: It also includes data preprocessing of thermal-mechanical coupling simulation data, specifically: Based on the final stress field and the final temperature field, the initial multi-channel three-dimensional input tensor is constructed, which is expressed as follows: in, represents the initial multi-channel three-dimensional input tensor, The final stress field The normal stress component of represents the shear stress component of the final stress field, represents the final temperature field; Normalizing the initial multi-channel three-dimensional input tensor and reducing the three-dimensional data size to obtain the final multi-channel three-dimensional input tensor; Based on the location set of high crack risk areas, a crack risk label tensor is constructed, which is expressed as follows: in, represents the crack risk label tensor, represents the location set of high crack risk areas, Indicates a location point; A sample set is formed based on the final multi-channel three-dimensional input tensor and the crack risk label tensor.

6. The method for controlling cracks in ceramic bearing rollers based on artificial intelligence according to claim 4 is characterized in that: The method of obtaining initial crack parameters based on the predicted high crack risk area location set comprises the following steps: Calculate the centroid position of the predicted high crack risk area location set to obtain the initial crack position, which is expressed as follows: in, The centroid of the predicted high crack risk area location set is the initial crack location. represents the predicted high crack risk area location set, The number of voxels representing the set of predicted high crack risk area locations; Fitting the initial crack shape according to the spatial distribution of the predicted high crack risk area location set to obtain initial crack phenotype parameters, wherein the initial crack phenotype parameters include initial crack shape and initial crack size; The maximum principal stress of the predicted high crack risk area location set is calculated, and then the average value of the maximum principal stress direction is obtained. The average value of the maximum principal stress direction is normalized to a unit vector, and the unit vector is the initial crack direction. The average value of the maximum principal stress direction is expressed as follows: in, represents the average value of the maximum principal stress direction of the predicted high crack risk area location set, is the eigenvector of the maximum principal stress of the predicted high crack risk area location set.

7. The method for controlling cracks in ceramic bearing rollers based on artificial intelligence according to claim 1, characterized in that: The method of obtaining crack propagation simulation information by using initial crack parameters, a comprehensive finite element model and material physical parameters comprises the following steps: The initial crack position is used as a reference system, and the initial crack phenotype parameters are matched to the comprehensive finite element model. Then, the crack surface normal vector of the comprehensive finite element model is matched to the initial crack direction and the mesh of the initial crack area is refined to update the comprehensive finite element model. The crack propagation condition, crack propagation direction and crack propagation increment are defined to obtain the crack propagation finite element model. Based on the crack propagation finite element model, the control equation is iteratively solved through spatial discretization and time discretization. Based on the data of each iteration in the solution process, the crack propagation simulation information is obtained. The solution process is the crack propagation simulation process, which specifically includes: When the crack propagation conditions are met, crack propagation is performed to obtain the current crack propagation speed, crack propagation increment, crack propagation direction, stress field and temperature field, and then the crack propagation finite element model is updated, wherein the crack propagation increment is obtained based on the crack propagation speed; Based on the updated crack propagation finite element model, the control equation is re-solved until the solution termination condition is met, and then the crack propagation simulation process is terminated; Based on the stress field and temperature field of each iteration, the maximum stress and maximum temperature are obtained.

8. The method for controlling cracks in ceramic bearing rollers based on artificial intelligence according to claim 7 is characterized in that: The crack extension condition includes that crack extension is performed if the current crack energy release rate is not less than the energy release threshold, and the energy release rate is expressed as follows: in, represents the current energy release rate, represents the equivalent elastic modulus of the ceramic bearing roller, represents the stress intensity factor of the opening mode of the previous iteration, represents the slip mode stress intensity factor of the previous iteration, represents the stress intensity factor of the tearing mode of the previous iteration, which are expressed as follows: in, represents the maximum principal stress in the normal direction of the current crack tip, represents the crack length of the previous iteration, represents the tangential shear stress of the stress tensor at the current crack tip, The normal shear stress represents the stress tensor at the current crack tip; The crack propagation direction is expressed as follows: in, represents the current crack propagation direction, represents the crack growth direction of the previous iteration, represents the crack extension angle of the previous iteration, ; The crack growth rate and the crack growth increment are expressed as follows: in, represents the current crack growth rate, represents the crack growth rate constant, Indicates the critical energy release rate of the material of the ceramic bearing roller, i.e. the energy release threshold. represents the crack growth index, represents the current crack growth increment, Indicates the time difference.

9. The method for controlling cracks in ceramic bearing rollers based on artificial intelligence according to claim 1, characterized in that: The method of constructing an optimization model based on the crack propagation simulation information and the equipment processing parameters and optimizing the equipment processing parameters to obtain the optimal processing parameters of the equipment includes the following steps: Constructing an optimization model, wherein the optimization model at least includes an optimization variable model, an optimization target model and a constraint condition model. The optimization variable model is constructed based on equipment processing parameters. The equipment processing parameters are optimization variables that at least include cutting speed, feed rate, cutting depth and coolant flow rate. The optimization target model is constructed based on a first crack risk value and a processing cost value. The first crack risk value is obtained based on crack extension simulation information. The processing cost value at least includes a processing time cost value, a tool wear cost value and an energy consumption cost value. The constraint condition model includes an equipment processing parameter constraint condition model and a processing quality constraint condition model. The processing quality constraint condition model is constructed based on surface roughness, processing dimensional accuracy and processing tool wear. The optimization model is preliminarily solved by genetic algorithm, and the optimal population is obtained as the preliminary optimization processing parameters. Among them, the population to be optimized in the genetic algorithm is the equipment processing parameters, the fitness of the population is obtained by optimizing the target model, and the encoding of the population is controlled by the constraint condition model; The optimization model is globally solved by the particle swarm algorithm, and the global optimal position of the particle swarm is the optimal processing parameter of the equipment. The initial position of the particle is the preliminary optimized processing parameter, the fitness of the particle is obtained by optimizing the target model, and the position of the particle is adjusted by the constraint condition model. The optimization variable model, the optimization target model, the equipment processing parameter constraint model, the processing quality constraint model, the first crack risk value and the processing cost value are respectively expressed as follows: in, represents the optimization variable, i.e., the equipment processing parameters, Indicates the cutting speed, Indicates the feed rate, Indicates the cutting depth, Indicates the coolant flow rate, represents the optimization goal, represents the weight coefficient, and , represents the first crack risk value, Indicates the processing cost value, Indicates the surface roughness of ceramic bearing rollers. Indicates the machining dimensional accuracy required for ceramic bearing rollers. Indicates the wear of the machining tool, Indicates the maximum and minimum values ​​of the corresponding parameters, represents the maximum stress, is the crack growth rate, Indicates the maximum temperature, Both represent weight coefficients, represents the processing time cost value, , represents the unit time cost, Indicates the length of the ceramic bearing roller, represents the tool wear cost value, , represents the unit tool wear cost value, represents the energy consumption cost value, , represents the unit energy cost, Indicates energy consumption.

10. The method for controlling cracks in ceramic bearing rollers based on artificial intelligence according to claim 1, characterized in that: The method of inferring the processing parameters of the current period based on the crack risk prediction model to obtain the crack risk prediction value of the current period, and adjusting the actual optimal processing parameters of the current period based on the crack risk prediction value of the current period, the optimal processing parameters of the current period and the processing parameters of the current period to achieve the control of the cracks of the ceramic bearing roller includes the following steps: Constructing a crack risk prediction pre-training model, and training the crack risk prediction pre-training model based on a historical first processing parameter of the equipment to obtain a crack risk prediction model, wherein the first processing parameter at least includes cutting force data, vibration data, temperature data, and a crack risk value; The first processing parameter of the equipment in the current period is inferred by the crack risk prediction model to obtain the crack risk prediction value in the current period; Based on the crack risk prediction value of the current period, the optimal processing parameters of the current period and the processing parameters of the current period, the adjusted processing parameters of the current period are obtained, specifically including: obtaining the risk error based on the crack risk prediction value of the current period and the preset target crack risk level, obtaining the adjustment amount of the theoretical processing parameters of the current period based on the risk error, the optimal processing parameters of the current period and the processing parameters of the current period, and then obtaining the theoretical optimal processing parameters of the current period, limiting the theoretical optimal processing parameters of the current period within a preset range, and obtaining the adjusted processing parameters of the current period; Adjusting the actual optimal processing parameters of the current period based on the adjusted processing parameters of the current period to achieve control of the cracks of the ceramic bearing roller; Among them, the risk error, the theoretical processing parameter adjustment amount in the current period, and the adjusted processing parameters in the current period are expressed as follows: in, represents the risk error, Indicates the preset target crack risk level, Indicates the crack risk prediction value of the current period, Indicates the theoretical processing parameter adjustment amount in the current period. represents the control gain, represents the weight coefficient, Indicates the optimal processing parameters for the current period, Indicates the processing parameters of the current period. Indicates the processing parameters adjusted in the current period. Indicates the theoretical optimal processing parameters for the current period, and Indicates the minimum and maximum values ​​preset for the processing parameters. Indicates that the target parameter is limited between the preset minimum and maximum values.

11. A ceramic bearing roller crack control system based on artificial intelligence, characterized in that: It includes a material parameter acquisition module, a finite element model acquisition module, a thermal data acquisition module, a crack parameter acquisition module, a crack extension information acquisition module, an optimal processing parameter acquisition module and an actual optimal processing parameter acquisition module; The material parameter acquisition module is used to acquire the material physical parameters of the ceramic bearing roller; The finite element model acquisition module is used to acquire a two-dimensional microscopic image sequence of the ceramic bearing roller and perform preprocessing, thereby constructing a three-dimensional microstructure digital twin model, performing finite element analysis on the three-dimensional microstructure digital twin model and combining it with material physical parameters to obtain an initial finite element model of the ceramic bearing roller; The thermal data acquisition module is used to apply mechanical boundary conditions and thermal boundary conditions to the initial finite element model to obtain a comprehensive finite element model, and to obtain thermal-mechanical coupling simulation data in combination with material physical parameters, wherein the thermal-mechanical coupling simulation data at least includes a final temperature field, a final stress field, and a set of locations of high crack risk areas; The crack parameter acquisition module is used to construct a crack initiation pre-training model and train it through thermal-mechanical coupling simulation data to obtain a crack initiation model, and then obtain initial crack parameters, wherein the initial crack parameters at least include initial crack position, initial crack direction and initial crack phenotype parameters; The crack extension information acquisition module is used to obtain crack extension simulation information through initial crack parameters, comprehensive finite element model and material physical parameters, wherein the crack extension simulation information at least includes crack extension speed, maximum stress and maximum temperature; The optimal processing parameter acquisition module is used to construct an optimization model and optimize the equipment processing parameters based on the crack propagation simulation information and the equipment processing parameters to obtain the optimal processing parameters of the equipment; The actual optimal processing parameter acquisition module is used to infer the processing parameters of the current period based on the crack risk prediction model to obtain the crack risk prediction value of the current period, and adjust the actual optimal processing parameters of the current period based on the crack risk prediction value of the current period, the optimal processing parameters of the current period and the processing parameters of the current period to achieve control of the cracks of the ceramic bearing roller; The step of applying mechanical boundary conditions and thermal boundary conditions to the initial finite element model to obtain a comprehensive finite element model and combining the physical parameters of the material to obtain thermal-mechanical coupling simulation data comprises the following steps: Applying mechanical boundary conditions to the initial finite element model to achieve displacement constraints, thereby obtaining a first finite element model; Applying thermal boundary conditions to the first finite element model to obtain a comprehensive finite element model, wherein the thermal boundary conditions include temperature boundary conditions and heat flow boundary conditions; Based on the comprehensive finite element model and material physical parameters, the thermal-mechanical coupling simulation data is obtained. The specific process includes: Constructing control equations, wherein the control equations include at least a mechanical equilibrium equation, a heat conduction equation, and a constitutive relationship equation, and solving the control equations by finite element discretization and time discretization in combination with a comprehensive finite element model to obtain a final temperature field, a final stress field, and a set of locations of high crack risk areas; The mechanical equilibrium equation, heat conduction equation and constitutive relationship equation are expressed as follows: in, represents stress, represents body density, represents the density of ceramic materials, represents displacement, Indicates time, represents the specific heat capacity of ceramic materials, Indicates temperature, represents the thermal conductivity of ceramic materials, represents the heat source term, represents the thermal conductivity term, represents the elastic stiffness matrix, represents the total strain, represents the thermal strain caused by temperature change, represents the matrix transpose, represents the thermal expansion coefficient of ceramic materials, represents the reference temperature, Represents the identity matrix.

12. A computer program product, characterized in that The method comprises computer instructions, which, when executed by a processor, implement the method according to any one of claims 1 to 10.

13. A ceramic bearing roller crack control device based on artificial intelligence, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that: When the processor executes the computer program, the method according to any one of claims 1 to 10 is implemented.

Images