A High-Precision Modeling Method for Porous Metallic Materials Based on Stochastic Algorithms and Its Application in Simulating the Electrothermal-Mechanical Properties of Materials

Abstract

This invention provides a high-precision modeling method for porous metallic materials based on a stochastic algorithm and its application in simulating the electrothermal-mechanical properties of materials. It relates to the field of artificial intelligence algorithm technology. The method includes: reconstructing and optimizing an initial reconstruction model of the porous metallic material based on a stochastic algorithm to obtain a three-dimensional reconstruction model; performing slice analysis on the three-dimensional reconstruction model and determining whether the analysis results meet the accuracy error standard; if they do, outputting the three-dimensional reconstruction model. By reconstructing and optimizing the initial model, the three-dimensional structure of porous materials can be accurately simulated, achieving high-precision modeling of porous metallic materials. Generating and growing pore seeds using a stochastic algorithm ensures the randomness and realism of the model. The setting of the accuracy error standard further improves the reliability of the model, ensuring the accuracy of porosity and pore size. This effectively solves the error problem in the modeling process of porous metallic materials, providing a solid foundation for high-precision simulation and material performance prediction.

Description

Technical Field

[0001] This invention relates to the field of artificial intelligence algorithm technology, and in particular to a high-precision modeling method for porous metallic materials based on random algorithms and its application in the simulation of the electrothermal-mechanical properties of materials. Background Technology

[0002] Currently, porous materials, due to their unique structural characteristics, have broad application prospects in fields such as electronic packaging, energy storage, aerospace, and biomedicine. In particular, porous structures with non-uniform, random pore distributions and a homogeneous or single-metal matrix exhibit macroscopic properties closely related to their microscopic pore characteristics, especially porosity and pore size distribution. These two key parameters directly affect the material's core properties such as electrical conductivity, thermal conductivity, mechanical strength, and dielectric properties. Therefore, accurate characterization and reproduction of porosity and pore size distribution are crucial for high-precision structural modeling.

[0003] Currently, commonly used methods for characterizing pore structure include scanning electron microscopy (SEM), X-ray computed tomography (X-CT), and mercury intrusion porosimetry (MIP). However, these methods are complex to operate, require expensive equipment, are time-consuming to analyze, and are difficult to perform in-situ real-time monitoring. Furthermore, existing numerical modeling methods are mostly based on idealized periodic or regular pore assumptions, making it difficult to realistically reproduce the non-uniform random pore characteristics of materials, resulting in significant discrepancies between simulation results and actual material properties. Simultaneously, the lack of accurate structure-property mapping models makes it difficult to accurately predict material properties under high-frequency, high-temperature, and complex loads, limiting the performance optimization and engineering applications of porous materials.

[0004] Therefore, there is an urgent need to propose a high-precision modeling method for randomly distributed pores, single or homogeneous systems, which can quickly reconstruct the real pore structure and effectively predict its electrical, thermal, mechanical and other properties, providing theoretical basis and computational support for the performance optimization and engineering design of porous materials. Summary of the Invention

[0005] One of the objectives of this invention is to provide a high-precision modeling method for porous metallic materials based on a random algorithm, in order to solve the problems in the background art.

[0006] This invention provides a high-precision modeling method for porous metallic materials based on a random algorithm, comprising:

[0007] S1. Based on a stochastic algorithm, the initial reconstruction model of porous metal material is reconstructed and optimized to obtain a three-dimensional reconstruction model.

[0008] S2. Slice and analyze the 3D reconstruction model, and determine whether the analysis results meet the accuracy error standard;

[0009] S3. If the conditions are met, output the 3D reconstruction model.

[0010] Optionally, the initial reconstruction model is generated based on the initial reconstruction parameters;

[0011] The initial reconstruction parameters include at least: defining the model size, the initial growth probability in each direction, the pore seed probability, and the initial volumetric porosity.

[0012] Optionally, the initial reconstruction and optimization steps of the model include:

[0013] S11. Based on the probability of the pore seed, several pore seeds are randomly generated in the initial reconstruction model; each pore seed grows randomly to its neighboring units.

[0014] S12. Assign a random number between 0 and 1 to each neighboring unit; if the random number assigned to a neighboring unit is less than the growth probability in the corresponding direction, then the corresponding neighboring unit is occupied by the porous phase.

[0015] S13. Repeat S11 to S12 until the initial reconstructed model reaches the initially set target volume porosity.

[0016] S14. Use a noise filter to smooth the pore edges of the initial reconstructed model to achieve the target volumetric porosity.

[0017] Optionally, the steps for slicing and analyzing the 3D reconstructed model include:

[0018] The 3D reconstruction model is uniformly cut into multiple 2D slices along a preset direction to match the cross-sectional sampling method during actual SEM observation.

[0019] The simulated average porosity and simulated average pore diameter of multiple two-dimensional slices were calculated and used as the analysis results.

[0020] Optional accuracy error standards include:

[0021] The analysis results show that the error between the simulated average porosity and simulated average pore size and the target average two-dimensional porosity and target average pore size is less than or equal to the threshold.

[0022] Optionally, the steps for obtaining the target average two-dimensional porosity and the target average pore size include:

[0023] Based on at least a predetermined number of scanning electron microscope images of the porous metal material, determine the target average two-dimensional porosity and the target average pore size.

[0024] Optionally, the noise filter may include at least: a kernel filter, a median filter, and a Gaussian filter.

[0025] Optionally, if the analysis results do not meet the accuracy error standard, adjust the initial reconstruction parameters of the initial reconstruction model, and then re-execute S1 to S3.

[0026] Optionally, the growth probability includes at least: isotropic and anisotropic.

[0027] The application of any of the methods provided in the embodiments of the present invention in the simulation of the electrothermal-mechanical properties of materials includes:

[0028] The output 3D reconstructed model is then used to simulate its electrothermal-mechanical properties using a finite element simulation software platform.

[0029] The finite element simulation software platforms include at least: Ansys, COMSOL Multiphysics, and Abaqus.

[0030] The present invention has achieved the following beneficial effects:

[0031] 1. This invention, through reconstruction and optimization of the initial model, can accurately simulate the three-dimensional structure of porous materials, achieving high-precision modeling of porous metallic materials. The generation and growth of pore seeds using a random algorithm ensures the randomness and realism of the model. The introduction of slice analysis verifies the accuracy of the three-dimensional model, ensuring consistency between simulation results and actual observations. The setting of accuracy error standards further improves the reliability of the model, guaranteeing the accuracy of porosity and pore size. Overall, this invention effectively solves the error problem in the modeling process of porous metallic materials, providing a solid foundation for high-precision simulation and material property prediction.

[0032] 2. This invention is applicable to porous homogeneous metallic materials with randomly distributed pore structures. It breaks through the technical limitations of traditional modeling based on regular periodic structures, and has stronger material adaptability and microstructure expression capabilities. It is especially suitable for engineering application scenarios with complex morphology and strong structural randomness, effectively expanding the applicable boundaries of this type of method in real material modeling.

[0033] 3. The modeling process of this invention is simple and efficient. It can reconstruct the three-dimensional pore structure using readily available two-dimensional SEM image information, eliminating the need for costly experimental methods such as X-ray tomography. This significantly reduces the technical threshold and time cost of model building, making it suitable for batch and automated modeling scenarios. Furthermore, this invention introduces porosity and pore size distribution parameters extracted from actual images as constraints, improving the model's accuracy in reproducing the real pore morphology and providing an accurate structural input basis for subsequent performance simulations.

[0034] 4. The three-dimensional porous structure model generated by this invention has good compatibility with finite element simulation and can be directly imported into multiple mainstream multiphysics analysis platforms to conduct thermal, electrical, and mechanical performance simulations, supporting structural response prediction and functional evaluation under complex working conditions. After structural correction and error feedback optimization, the simulation results show a high degree of consistency with actual test data, with the error controlled within an acceptable range (5%). It can be widely used in material structure-property relationship research, performance prediction, and engineering reliability assessment, with good accuracy assurance and application value.

[0035] 5. This invention realizes the linkage and integration between scanning electron microscope (SEM) image recognition, three-dimensional pore structure reconstruction and finite element simulation, and establishes an integrated modeling process of image input, random modeling and performance prediction, overcoming the problems of inconsistent data interfaces and scattered processing steps in traditional methods.

[0036] Other features and advantages of the invention will be set forth in the following description, and will be apparent in part from the description, or may be learned by practicing the invention. The objects and other advantages of the invention may be realized and obtained by means of the structures particularly pointed out in the written description and the accompanying drawings.

[0037] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0038] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings:

[0039] Figure 1 This is a schematic diagram of a high-precision modeling method for porous metallic materials based on a random algorithm, as described in an embodiment of the present invention.

[0040] Figure 2 This is a schematic diagram of the growth direction of pores in three-dimensional space in an embodiment of the present invention;

[0041] Figure 3 This is a schematic diagram of SEM images of sintered joints under different pore gradients in an embodiment of the present invention;

[0042] Figure 4 This is a schematic diagram showing the results of thermoelectric performance simulation under different pore gradients and joint simulation thermal / electrical conductivity under different pore gradients in the embodiments of the present invention.

[0043] Figure 5 This is another schematic diagram showing the simulation results of thermoelectric performance under different pore gradients and the simulated thermal / electrical conductivity of the joint under different pore gradients in the embodiments of the present invention. Detailed Implementation

[0044] The preferred embodiments of the present invention will be described below with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are for illustration and explanation only and are not intended to limit the present invention.

[0045] Example 1:

[0046] This invention provides a high-precision modeling method for porous metallic materials based on a random algorithm, such as... Figure 1 As shown, it includes:

[0047] S1. Based on a stochastic algorithm, the initial reconstruction model of porous metal material is reconstructed and optimized to obtain a three-dimensional reconstruction model.

[0048] S2. Slice and analyze the 3D reconstruction model, and determine whether the analysis results meet the accuracy error standard;

[0049] S3. If the conditions are met, output the 3D reconstruction model.

[0050] The initial reconstruction model is generated based on the initial reconstruction parameters;

[0051] The initial reconstruction parameters include at least: defining the model size, the initial growth probability in each direction, the pore seed probability, and the initial volume porosity.

[0052] The initial reconstruction and optimization steps of the model include:

[0053] S11. Based on the probability of the pore seed, several pore seeds are randomly generated in the initial reconstruction model; each pore seed grows randomly to its neighboring units.

[0054] S12. Assign a random number between 0 and 1 to each neighboring unit; if the random number assigned to a neighboring unit is less than the growth probability in the corresponding direction, then the corresponding neighboring unit is occupied by the porous phase.

[0055] S13. Repeat S11 to S12 until the initial reconstructed model reaches the initially set target volume porosity.

[0056] S14. Use a noise filter to smooth the pore edges of the initial reconstructed model to achieve the target volume porosity;

[0057] The steps involved in slicing and analyzing a 3D reconstructed model include:

[0058] The 3D reconstruction model is uniformly cut into multiple 2D slices along a preset direction to match the cross-sectional sampling method during actual SEM observation.

[0059] The simulated average porosity and simulated average pore size of multiple two-dimensional slices were calculated and used as the analysis results.

[0060] Accuracy error standards include:

[0061] The error between the simulated average porosity and simulated average pore size and the target average two-dimensional porosity and target average pore size in the analysis results is less than or equal to the threshold.

[0062] Growth probability includes at least: isotropic and anisotropic.

[0063] The working principle and beneficial effects of the above technical solution are as follows:

[0064] Input the initial reconstruction parameters, including defining the model size (lx, ly, lz) and the initial growth probabilities G in each direction. i The probability S of pore seeds p and initial volumetric porosity P 3D-init An initial reconstruction model is generated based on the initial reconstruction parameters.

[0065] Based on the pore seed probability S p In the initial reconstruction model, several pore seeds are randomly generated, and each pore seed grows randomly towards neighboring cells. Each neighboring cell is assigned a random number between 0 and 1. If this value is less than the growth probability G in the corresponding direction... i Then the unit is occupied by the porous phase. This process is repeated until the initially set target volumetric porosity P is reached. 3D-init Subsequently, a noise filter was used to smooth the pore edges, improving the structural quality. The pore growth direction is as follows: Figure 2 As shown.

[0066] The pore seed probability is a probability parameter that determines the distribution of pore seeds (initial pore positions). It defines the frequency or probability of generating pore seeds in the initial reconstruction model and determines the likelihood of each position becoming a pore seed.

[0067] A pore seed is the starting point or seed in a model; it is the initial location of a pore that gradually expands to neighboring cells through random growth in subsequent steps.

[0068] Neighboring cells refer to cells or regions directly adjacent to the current pore seed location. During growth, pores expand into these neighboring cells. The definition of a neighboring cell can be geometrically adjacent or determined by the model's mesh generation.

[0069] The random number is a number between 0 and 1 used to determine whether pores are allowed to expand into a neighboring cell. This random number is compared with the growth probability to determine whether to occupy the neighboring cell.

[0070] The growth probability is the probability that a pore seed will expand into a neighboring cell. Each direction or each neighboring cell has a growth-related probability value. If the generated random number is less than this growth probability, the pore will expand into that neighboring cell.

[0071] The target volumetric porosity is the design objective of the reconstructed model, which is the proportion of pore volume to the total model volume in the final model. The model will continuously grow pores until the set volumetric porosity is reached.

[0072] The 3D reconstruction model is uniformly cut into multiple 2D images along a preset direction to match the cross-sectional sampling method during actual SEM observation. The simulated average porosity P of the multiple 2D slices is calculated respectively. 2D-sim and simulated average aperture A 2D-sim .

[0073] If the simulated average porosity P of multiple two-dimensional slices 2D-sim and simulated average aperture A 2D-sim With the target average two-dimensional porosity P 2D-target With the target average aperture A 2D-target Error (simulated average porosity P) 2D-sim With the target average two-dimensional porosity P 2D-target Error, simulated average aperture A 2D-sim With the target average aperture A 2D-target If all errors are less than or equal to the threshold, the final 3D reconstruction model is output. The threshold can be set to 5%.

[0074] The growth probability can be set to isotropic (the probability is the same in all directions) or anisotropic (the probability is different in different directions) according to actual needs.

[0075] In each two-dimensional slice, the proportion of pores within that slice is calculated to obtain the porosity of that slice. Then, the porosity of all two-dimensional slices is averaged to obtain the simulation average porosity, which reflects the porosity of the entire model.

[0076] In each 2D slice, the pore diameter of all pores is calculated. Then, the average pore diameter of all 2D slices is taken to obtain the simulated average pore diameter, which reflects the size characteristics of the pores. The number of 2D slices can be flexibly determined according to the model size, structural complexity, and accuracy requirements.

[0077] This invention is applicable to porous materials in which the matrix material is a homogeneous metal or a single metal material system and the pore structure is non-uniformly and randomly distributed.

[0078] This invention, through the reconstruction and optimization of the initial model, can accurately simulate the three-dimensional structure of porous materials, achieving high-precision modeling of porous metallic materials. The generation and growth of pore seeds using a random algorithm ensures the randomness and realism of the model. The introduction of slice analysis verifies the accuracy of the three-dimensional model, ensuring consistency between simulation results and actual observations. The setting of accuracy error standards further improves the reliability of the model, guaranteeing the accuracy of porosity and pore size. Overall, this invention effectively solves the error problem in the modeling process of porous metallic materials, providing a solid foundation for high-precision simulation and material property prediction.

[0079] Example 2:

[0080] In this embodiment of the invention, the steps for obtaining the target average two-dimensional porosity and the target average pore size include:

[0081] Based on at least a predetermined number of scanning electron microscope images of the porous metal material, determine the target average two-dimensional porosity and the target average pore size.

[0082] Two-dimensional porosity and pore size are extracted from at least three scanning electron microscope (SEM) images to determine the target average two-dimensional porosity P. 2D-target With the target average aperture A 2D-target Three or more SEM images are used to improve the accuracy of pore structure feature analysis. The target two-dimensional average porosity obtained from the SEM images should be controlled within the range of 0% to 50%. When the porosity exceeds this range, problems such as high connectivity or unclear boundaries between pores may occur, which can easily lead to distortion of the pore topology or a decrease in the convergence of the reconstruction algorithm during model construction, thereby affecting the structural accuracy and simulation adaptability of the final model.

[0083] The target average two-dimensional porosity is obtained by extracting two-dimensional pore data from at least three SEM images, calculating the porosity of each image, and then taking the average porosity of these images.

[0084] By analyzing the pore size in at least three SEM images, the size data of each pore is extracted, the average pore diameter of each image is calculated, and the average value of the pore diameters of these images is taken to obtain the target average pore diameter.

[0085] Example 3:

[0086] In this embodiment of the invention, the noise filter includes at least: a kernel filter, a median filter, and a Gaussian filter.

[0087] Noise filters include, but are not limited to, kernel filters, median filters, and Gaussian filters.

[0088] Example 4:

[0089] In this embodiment of the invention, if the analysis results do not meet the accuracy error standard, the initial reconstruction parameters of the initial reconstruction model are adjusted, and S1 to S3 are re-executed after adjustment.

[0090] If the error is greater than 5%, return to step S1 to readjust the modeling parameters.

[0091] Example 5:

[0092] The application of any of the methods provided in the embodiments of the present invention in the simulation of the electrothermal-mechanical properties of materials includes:

[0093] The output 3D reconstructed model is then used to simulate its electrothermal-mechanical properties using a finite element simulation software platform.

[0094] The finite element simulation software platforms include at least: Ansys, COMSOL Multiphysics, and Abaqus.

[0095] The reconstructed three-dimensional porous model is imported into finite element analysis software for subsequent electro-thermal and mechanical multi-physics performance simulation analysis.

[0096] Finite element simulation analysis software includes, but is not limited to, Ansys, COMSOL Multiphysics, and Abaqus. In practical applications, relevant process parameters (such as drying temperature and time) can be adjusted according to the characteristics of the organic carrier system used in the preparation of the porous material to obtain the optimal pore structure and performance.

[0097] Example 6:

[0098] To obtain the pore structure characteristics in actual experiments, experiments were conducted on copper sintered materials at a sintering temperature of 250℃ and a sintering pressure of 20MPa for sintering times of 10 minutes, 20 minutes, and 30 minutes. The pore structure results are as follows: Figure 3 As shown. The average porosities of the three groups of samples, measured using ImageJ software, were 30.75%, 23.02%, and 16.50%, respectively, and were labeled as porosity classes A, B, and C. Correspondingly, the average pore sizes of the three groups of samples, calculated using MATLAB (2023a), were 0.0074 μm. 2 0.0090μm 2 and 0.0105μm 2 Subsequently, the high-precision modeling method based on a stochastic algorithm proposed in this paper is used to reconstruct the pore structure in three dimensions. Taking the reconstruction process of pore level A as an example, the scaling ratio between the actual size and the simulation model is set to... That is, the size of the simulation model is 10 times the actual SEM observation aperture. The specific steps are as follows:

[0099] Step S1: Input model size as 40×40×40 voxels. Based on the actual measured porosity, set the initial porosity to 35% (slightly higher than the actual SEM measurement of 30.75%, and the algorithm will then search in a decreasing manner), and set the pore growth probability to anisotropy.

[0100] Step S2: Based on three scanning electron microscope (SEM) cross-sectional images, the target's average two-dimensional porosity was determined to be 30.75%, and the target's average pore size was determined to be 0.0074 μm. 2 .

[0101] Step S3: Based on the set pore seed probability, several pore seeds are randomly generated inside the model; the seeds grow randomly to neighboring units according to the set anisotropic growth probability until the initial volume porosity set in step S1 is reached; then, a Kernel noise filter is used to smooth the pore edges to improve the realism and quality of the structure.

[0102] Step S4: The reconstructed 3D model is uniformly cut into ten 2D images along the Y direction to simulate the cross-sectional sampling method during actual SEM observation, and the average porosity and average pore size of each 2D slice are calculated.

[0103] Step S5: Compare and analyze the simulation results with the experimental data. Calculate the simulated average two-dimensional porosity and simulated average pore size of the ten two-dimensional slices, and compare them with the target average two-dimensional porosity and target average pore size determined in Step S2. If the error exceeds 5%, return to Step S1, readjust the initial modeling parameters, and proceed to Step S6 until the error meets the requirement of being within 5%.

[0104] Step S6: Output the final high-precision three-dimensional pore reconstruction model and save the corresponding three-dimensional pore structure parameters for subsequent simulation verification.

[0105] Step S7: Import the completed three-dimensional porous structure model into the finite element simulation platform Comsol to perform numerical simulation analysis of the material's thermal and electrical properties.

[0106] The modeling process for pore grades B and C is the same as described above. The simulated electrical and thermal properties of the material under different pore grades are as follows: Figure 4 , Figure 5 As shown in Table 1, the simulation results were corrected by introducing appropriate correction factors, including scaling factor (F) and offset factor (ΔK), and the final predicted thermal conductivity and electrical conductivity are shown in Table 2.

[0107] The results show that the present invention achieves accurate prediction of the electrothermal properties of materials within an error range of less than 5%. Table 1: Average temperature and potential of the upper and lower interfaces of the simulation model with different pore gradients, and the corresponding temperature difference and potential difference between the upper and lower interfaces.

[0108]

[0109] Table 2: Comparison of measured and simulated results of joint thermal / electrical conductivity under different pore gradients

[0110]

[0111]

[0112] Example 7:

[0113] In this embodiment of the invention, when outputting the three-dimensional reconstruction model, the user's depth modeling requirements information is obtained through emotional intervention, and the initial reconstruction parameters of the initial reconstruction model are adjusted accordingly based on the depth modeling requirements. After adjustment, S1 to S3 are re-executed.

[0114] The steps involved in the emotionally-driven acquisition of information required for deep modeling include:

[0115] When a user views the output 3D reconstruction model, if it is detected that multiple depth modeling requirements that the user hesitates to select have reached the upper limit of hesitation selection, multiple hesitation resolution mechanisms and their resolution timings are planned for the multiple depth modeling requirements.

[0116] When any resolution opportunity is reached in the future, the corresponding hesitation resolution mechanism will be implemented for the user;

[0117] After each hesitation resolution mechanism has been executed for the user, based on at least one hesitation resolution mechanism whose resolution time has been reached, the resolution time that has not been reached is optimized.

[0118] Based on the user's final selection of deep modeling requirements, the deep modeling requirement information is determined.

[0119] The working principle and beneficial effects of the above technical solution are as follows:

[0120] The need for deep modeling typically stems from high requirements for model accuracy, porosity, pore size distribution, and the precise reproduction of complex microstructures. These demands drive the continuous optimization and refinement of 3D reconstruction models to ensure that the models accurately reflect the properties and characteristics of materials in practical applications. However, when choosing deep modeling solutions, users may hesitate due to a lack of sufficient understanding of modeling methods, inadequate consideration of data processing and computational resources, and unclear expectations for the final application results. This hesitation reduces selection efficiency and hinders the rapid advancement of further optimization and refinement of the 3D reconstruction model.

[0121] Based on this, embodiments of the present invention plan multiple hesitation resolution mechanisms and their timing for multiple deep modeling requirement items that users hesitate to select. When the timing for resolution is reached in the future, the corresponding hesitation resolution mechanism is executed for the user to help resolve the hesitation in selecting the deep modeling requirement item. Finally, based on the deep modeling requirement item finally selected by the user, the deep modeling requirement information is determined, realizing the emotional intervention of the acquisition of deep modeling requirement information, improving the efficiency of the user's selection of deep modeling requirements, and can quickly promote the further optimization and refinement of the 3D reconstruction model.

[0122] Furthermore, after each hesitant resolution mechanism has been executed for a user, based on at least one hesitant resolution mechanism whose resolution time has been reached, the resolution time that has not been reached is optimized to fully ensure the suitability of executing hesitant resolution mechanisms for users and further improve the efficiency of their selection of deep modeling requirements.

[0123] Example 8:

[0124] In this embodiment of the invention, the step of identifying multiple depth modeling requirements that a user hesitates to select includes:

[0125] Based on the user's viewing behavior information when viewing the output 3D reconstruction model, the required depth modeling can be predicted in real time.

[0126] Establish a demand selection area. Whenever a predicted candidate depth modeling demand is selected by a user, the corresponding candidate depth modeling demand is placed in the demand selection area.

[0127] Whenever a new candidate depth modeling requirement is predicted and the user has not confirmed its selection, the candidate requirement area is displayed to the user, and the user is asked to remove the target candidate depth modeling requirement from the candidate requirement area.

[0128] If the user refuses, the target candidate depth modeling requirement will be treated as the depth modeling requirement item that the user hesitates to select.

[0129] The steps for determining the target candidate depth modeling requirements include:

[0130] Select the candidate depth modeling requirements in the candidate demand area whose discrepancy with the predicted new candidate depth modeling requirements exceeds a threshold as the target candidate depth modeling requirements;

[0131] In addition, the candidate depth modeling requirements already placed in the candidate depth modeling area and the candidate depth modeling requirements predicted to be new candidate depth modeling requirements are used as the target candidate depth modeling requirements.

[0132] The working principle and beneficial effects of the above technical solution are as follows:

[0133] User viewing behavior information when viewing the output 3D reconstruction model includes at least: viewpoint adjustment, selected area, interactive operations, viewed content, viewing time, and annotations and markings. This viewing behavior information reflects how the user wants to further optimize and refine the 3D reconstruction model; therefore, it can be used to predict potential depth modeling needs in real time. During prediction, an artificial intelligence model trained using a large amount of historical viewing behavior information labeled with potential depth modeling needs as training samples can be used. The degree of conflict refers to the degree of conflict between the depth modeling goals of potential depth modeling needs in the demand candidate area and the predicted new potential depth modeling needs. There is a removal indicator for both the potential depth modeling needs already placed in the demand candidate area and the predicted new depth modeling needs, indicating which potential depth modeling need needs to be removed due to conflict or other reasons if a new potential depth modeling need is predicted and placed in the demand candidate area. For example, suppose there is already a heat conduction modeling need in the demand candidate area, which simulates the thermal conductivity of porous metal at room temperature. The new prediction requirement is to simulate the thermal conduction behavior of porous metals under high-temperature conditions. These two requirements conflict in terms of modeling methods, simulation scope, and parameter settings. Therefore, the system will issue a removal instruction, requiring the removal of the original room-temperature thermal conduction requirement when the new high-temperature thermal conduction modeling requirement is added to the requirement selection area.

[0134] Whenever a new candidate depth modeling requirement is predicted and the user has not confirmed its selection, the system will also request the user to remove the target candidate depth modeling requirement from the requirement selection area. If the user refuses, it means that the target candidate depth modeling requirement is a depth modeling requirement that the user is hesitant to select.

[0135] In this embodiment of the invention, when identifying multiple deep modeling requirements that a user is hesitant to select, a requirement selection area is established. Whenever a predicted candidate deep modeling requirement is confirmed by the user, the corresponding candidate deep modeling requirement is placed in the requirement selection area. Based on the degree of contradiction and the removal instruction, a target candidate deep modeling requirement is determined, and the user is requested to remove the target candidate deep modeling requirement from the requirement selection area. If the user refuses, the target candidate deep modeling requirement is taken as the deep modeling requirement that the user is hesitant to select. This greatly improves the accuracy, comprehensiveness, and efficiency of identifying multiple deep modeling requirements that the user is hesitant to select, enhances the applicability of the system, and further improves the accuracy of subsequent hesitancy resolution.

[0136] Example 9:

[0137] In this embodiment of the invention, the step of obtaining the upper limit of hesitant selection includes:

[0138] The total time it takes for users to view the output 3D reconstruction model and the number of times users have resolved historical hesitations represent the upper limit of the user's tolerance for hesitation, thus obtaining the tolerance value;

[0139] Based on the allowable value, match the corresponding hesitant choice upper limit from the hesitant choice upper limit table.

[0140] The longer the total time a user spends viewing the output 3D reconstruction model, the less time they have left to hesitate in order to expedite further optimization and refinement of the model; thus, the lower the upper limit of the user's tolerance for hesitation. Conversely, a higher number of historical hesitation resolutions indicates more past choices by the user, further reducing the time available for hesitation and thus the lower the upper limit of the user's tolerance for hesitation. The final tolerance value can be calculated using the following weighted formula:

[0141] C = w1 × T + w2 × H; where C is the allowable value, T is the total time for the user to view the output 3D reconstruction model, H is the number of times the user has resolved historical hesitations, and w1 and w2 are preset weight values.

[0142] The hesitant selection upper limit table has different allowable values ​​corresponding to the hesitant selection upper limit. The hesitant selection upper limit can be the number of multiple deep modeling requirement items to be hesitant to select, or the type set of multiple deep modeling requirement items to be hesitant to select.

[0143] In this embodiment of the invention, when obtaining the upper limit of hesitation selection, the total time for the user to view the output 3D reconstruction model and the user's historical hesitation resolution count represent the upper limit of the user's allowable hesitation, and obtain the allowable value. Based on this, the corresponding upper limit of hesitation selection is matched from the hesitation selection upper limit table, which improves the accuracy and efficiency of obtaining the upper limit of hesitation selection, improves the accuracy of its use as multiple hesitation resolution mechanisms and the timing of their resolution planning, and improves the system's working efficiency.

[0144] Example 10:

[0145] In this embodiment of the invention, the planning steps for multiple hesitation resolution mechanisms and their timing include:

[0146] Based on multiple deep modeling requirements for hesitant choices, and matching hesitancy resolution experience;

[0147] Experiences in resolving hesitation and adapting to the viewing behavior scenarios of users who will view the predicted 3D reconstructed model of the output in the future;

[0148] Based on the experience of adapting to resolve hesitation, a hesitation resolution mechanism is determined, and the timing of the resolution is when the user enters the viewing behavior scenario adapted to the experience of adapting to resolve hesitation.

[0149] Hesitation resolution experience consists of pre-set experiences for resolving hesitation choices applicable to multiple deep modeling requirements. It predicts the user's future viewing behavior scenarios when viewing the output 3D reconstructed model, which can include content scenes within the output 3D reconstructed model. If the hesitation resolution experience fits the viewing behavior scenario, a hesitation resolution mechanism is determined based on the fitted experience (directly planned against the experience). The resolution timing is when the user enters the viewing behavior scenario adapted to that fitted hesitation resolution experience. This improves the accuracy and efficiency of multiple hesitation resolution mechanisms and their timing planning, significantly enhancing its effectiveness in resolving user hesitation choices.

[0150] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A high-precision modeling method for porous metallic materials based on a stochastic algorithm, characterized in that, include: S1. Based on a stochastic algorithm, the initial reconstruction model of porous metal material is reconstructed and optimized to obtain a three-dimensional reconstruction model. S2. Slice and analyze the 3D reconstruction model, and determine whether the analysis results meet the accuracy error standard; S3. If the conditions are met, output the 3D reconstruction model; When outputting the 3D reconstruction model, the system emotionally intervenes to obtain the user's deep modeling needs information, and adjusts the initial reconstruction parameters of the initial reconstruction model accordingly based on the deep modeling needs. After adjustment, S1 to S3 are re-executed. The steps involved in the emotionally-driven acquisition of information required for deep modeling include: When a user views the output 3D reconstruction model, if it is detected that multiple depth modeling requirements that the user hesitates to select have reached the upper limit of hesitation selection, multiple hesitation resolution mechanisms and their resolution timings are planned for the multiple depth modeling requirements. When any resolution opportunity is reached in the future, the corresponding hesitation resolution mechanism will be implemented for the user; After each hesitation resolution mechanism completes its execution for the user, based on at least one hesitation resolution mechanism whose resolution time has been reached, optimization is performed on the resolution time that has not been reached. Based on the user's final selection of deep modeling requirements, determine the deep modeling requirements information; The steps for identifying multiple deep modeling requirements that a user is hesitant to select include: Based on the user's viewing behavior information when viewing the output 3D reconstruction model, the required depth modeling can be predicted in real time. Establish a demand selection area. Whenever a predicted candidate depth modeling demand is selected by a user, the corresponding candidate depth modeling demand is placed in the demand selection area. Whenever a new candidate depth modeling requirement is predicted and the user has not confirmed its selection, the candidate requirement area is displayed to the user, and the user is asked to remove the target candidate depth modeling requirement from the candidate requirement area. If the user refuses, the target candidate depth modeling requirement will be treated as the depth modeling requirement item that the user hesitates to select. The steps for determining the target candidate depth modeling requirements include: Select the candidate depth modeling requirements in the candidate demand area whose discrepancy with the predicted new candidate depth modeling requirements exceeds a threshold as the target candidate depth modeling requirements; In addition, the candidate depth modeling requirements already placed in the candidate depth modeling area and the candidate depth modeling requirements predicted to be new candidate depth modeling requirements are used as the target candidate depth modeling requirements.

2. The high-precision modeling method for porous metallic materials based on a random algorithm as described in claim 1, characterized in that, The initial reconstruction model is generated based on the initial reconstruction parameters; The initial reconstruction parameters include at least: defining the model size, the initial growth probability in each direction, the pore seed probability, and the initial volumetric porosity.

3. The high-precision modeling method for porous metallic materials based on a random algorithm as described in claim 2, characterized in that, The initial reconstruction and optimization steps of the model include: S11. Based on the probability of the pore seed, several pore seeds are randomly generated in the initial reconstruction model; each pore seed grows randomly to its neighboring units. S12. Assign a random number between 0 and 1 to each neighboring unit; if the random number assigned to a neighboring unit is less than the growth probability in the corresponding direction, then the corresponding neighboring unit is occupied by the porous phase. S13. Repeat S11 to S12 until the initial reconstructed model reaches the initially set target volume porosity. S14. Use a noise filter to smooth the pore edges of the initial reconstructed model to achieve the target volumetric porosity.

4. The high-precision modeling method for porous metallic materials based on a random algorithm as described in claim 1, characterized in that, The steps involved in slicing and analyzing a 3D reconstructed model include: The 3D reconstruction model is uniformly cut into multiple 2D slices along a preset direction to match the cross-sectional sampling method during actual SEM observation. The simulated average porosity and simulated average pore diameter of multiple two-dimensional slices were calculated and used as the analysis results.

5. The high-precision modeling method for porous metallic materials based on a random algorithm as described in claim 1, characterized in that, Accuracy error standards include: The analysis results show that the error between the simulated average porosity and simulated average pore size and the target average two-dimensional porosity and target average pore size is less than or equal to the threshold.

6. The high-precision modeling method for porous metallic materials based on a random algorithm as described in claim 5, characterized in that, The steps for obtaining the target average two-dimensional porosity and the target average pore size include: Based on at least a predetermined number of scanning electron microscope images of the porous metal material, determine the target average two-dimensional porosity and the target average pore size.

7. The high-precision modeling method for porous metallic materials based on a random algorithm as described in claim 3, characterized in that, Noise filters include at least: kernel filters, median filters, and Gaussian filters.

8. The high-precision modeling method for porous metallic materials based on a random algorithm as described in claim 2, characterized in that, If the analysis results do not meet the accuracy error standard, adjust the initial reconstruction parameters of the initial reconstruction model, and then re-execute S1 to S3.

9. The high-precision modeling method for porous metallic materials based on a random algorithm as described in claim 2, characterized in that, Growth probability includes at least: isotropic and anisotropic.

10. The application of the method according to any one of claims 1 to 9 in the simulation of the electrothermal-mechanical properties of materials, characterized in that, include: The output 3D reconstructed model is then used to simulate its electrothermal-mechanical properties using a finite element simulation software platform. The finite element simulation software platforms include at least: Ansys, COMSOL Multiphysics, and Abaqus.

Images