A method and system for sintering simulation homogenization of ceramic lattice structures

Abstract

The application provides a sintering simulation homogenization method and system for a ceramic lattice structure, and belongs to the technical field of ceramic additive manufacturing and numerical simulation; the method comprises the following steps: constructing a lattice structure model and an equivalent homogeneous structure model, and obtaining macroscopic displacement time sequence data; simulating the evolution process of the macroscopic displacement based on the constructed sintering process proxy model; pre-training the sintering process proxy model by using a joint loss function with physical constraint conditions; performing global optimization based on a differential evolution algorithm, and inversely solving equivalent constitutive parameters of the homogeneous structure, so that the displacement time-varying curves of the homogeneous structure and the ceramic lattice structure are consistent under the same sintering process. The application takes ceramic sintering displacement as a core index of homogenization equivalence, can significantly reduce the trial-and-error cost and calculation resource consumption of a large ceramic lattice structure in a process development process, and further provides effective theoretical guidance for ceramic additive manufacturing.

Description

Technical Field

[0001] This invention belongs to the field of ceramic additive manufacturing and numerical simulation technology, and particularly relates to a method and system for homogenization of sintering simulation for ceramic lattice structural parts. Background Technology

[0002] The statements in this section are merely background information related to the present invention and do not necessarily constitute prior art.

[0003] Ceramic lattice structures, due to their high specific strength, controllable porosity, and excellent energy absorption performance, have broad application prospects in aerospace, biomedicine, and energy equipment. In current engineering practice, determining sintering process parameters relies heavily on repeated trial-and-error experiments. Each sintering experiment not only consumes expensive ceramic powder raw materials and energy but also has a long cycle; the process finalization of a single complex component often takes weeks or even months, resulting in high R&D costs and severely restricting the feasibility of moving ceramic lattice structures from the laboratory to large-scale production. To reduce experimental costs and improve process predictability, numerical simulation has become an important tool in sintering process design.

[0004] However, existing sintering simulation methods generally suffer from the following technical shortcomings: (1) Existing sintering simulation methods are mainly based on finite element models, in which the creep constitutive relations are often constructed using the SOVS (Skorohold-Olevsky Viscous Sintering) viscous flow model, which includes multiple parameters to be fitted. Based on this, for lattice structures, the cost of directly performing high-fidelity sintering simulation calculations is extremely high.

[0005] (2) Existing homogenization methods for sintering simulation are mostly applicable to linear elastic materials, which are difficult to handle the strong nonlinearity and time-dependent densification behavior in the sintering process. At the same time, physical information neural networks based on partial differential equations often cause gradient explosion or get stuck in local optima when inverting highly nonlinear sintering parameters due to high-order derivatives. Summary of the Invention

[0006] To overcome the shortcomings of the prior art, this invention provides a sintering simulation homogenization method and system for ceramic lattice structures. Using ceramic sintering displacement as the core indicator of homogenization equivalence, it can significantly reduce the trial-and-error costs and computational resource consumption of large ceramic lattice structures in the process development, thereby providing effective theoretical guidance for the manufacturing of ceramic additive materials.

[0007] To achieve the above objectives, one or more embodiments of the present invention provide the following technical solutions: The first aspect of this invention provides a method for homogenization in sintering simulation of ceramic lattice structures.

[0008] A method for homogenization in sintering simulation of ceramic lattice structures, comprising: Construct a lattice structure model and an equivalent homogeneous structure model to obtain macroscopic displacement time series data of the ceramic lattice structure and homogeneous structure during the sintering simulation process; A sintering process proxy model based on a long short-term memory network is constructed to simulate the evolution of macroscopic displacement according to given constitutive parameters and temperature conditions. The sintering process proxy model is pre-trained based on a joint loss function with physical constraints. Global optimization is performed based on the differential evolution algorithm, and the equivalent constitutive parameters of the homogeneous structure are solved by inversion, so that the displacement time-varying curves of the homogeneous structure and the ceramic lattice structure are consistent under the same sintering process.

[0009] Furthermore, the SOVS viscous flow model is used as the material constitutive model to construct a lattice structure model with fine porosity and a solid equivalent homogeneous structure model; wherein, the SOVS viscous flow model is expressed as: ; In the formula, Indicates inelastic strain rate; Indicates the fully dense shear modulus. Indicates the current temperature; Represents the normalized shear modulus. Represents the normalized bulk modulus. Indicates sintering stress; Represents relative density. Represents the deviatoric stress tensor; Represents the stress tensor traces, For Kronecker symbol.

[0010] Furthermore, the construction of the sintering process proxy model is achieved by sequentially performing input encoding, temporal feature extraction, physical state decoupling, activation temperature gating, and displacement accumulation operations.

[0011] Furthermore, the activation of temperature gating is achieved through a preset temperature gating mechanism, the calculation formulas for the gating values ​​during the contraction and thermal expansion phases of which are respectively expressed as: ; ; in, The current temperature. The sintering activation temperature, These are the gating parameters used to control smoothness; and These represent the gate values ​​for the contraction and thermal expansion phases, respectively.

[0012] Furthermore, the joint loss function Represented as: ; in, and Represent X Axial direction and Z Displacement loss in the axial direction, This represents a physical criterion penalty term, which is used to limit... Z The displacement in the direction is always greater than X The displacement in the direction of gravity is such that the settlement contraction in the direction of gravity is greater than the contraction in the horizontal direction.

[0013] Furthermore, global optimization is performed based on the differential evolution algorithm to inversely solve for the equivalent constitutive parameters of the homogeneous structure, including: The weights of the pre-trained sintering process surrogate model are frozen and used as a forward fast solver. Combined with the differential evolution global optimization algorithm, the population is initialized within the preset constitutive parameter boundary. The mean square error between the output of the sintering process surrogate model and the displacement data of the obtained ceramic lattice structure is used as the fitness function. Through crossover, mutation and selection operations, iterative evolution is carried out to search for the equivalent constitutive parameter combination that minimizes the fitness function.

[0014] Furthermore, the ceramic lattice structure is a three-period minimal surface structure, specifically represented as a Gyroid-type cell structure.

[0015] A second aspect of the present invention provides a sintering simulation homogenization system for ceramic lattice structural components.

[0016] A sintering simulation homogenization system for ceramic lattice structural components includes: The structure building module is configured to: build a lattice structure model and an equivalent homogeneous structure model, and obtain macroscopic displacement time series data of the ceramic lattice structure and homogeneous structure during the sintering simulation process; The model building module is configured to: build a sintering process proxy model based on a long short-term memory network, which is used to simulate the evolution of macroscopic displacement based on given constitutive parameters and temperature conditions; The pre-training module is configured to pre-train the sintering process proxy model based on a joint loss function with introduced physical constraints. The simulation inversion module is configured to perform global optimization based on the differential evolution algorithm, and invert the solution of the equivalent constitutive parameters of the homogeneous structure so that the displacement time-varying curves of the homogeneous structure and the ceramic lattice structure are consistent under the same sintering process.

[0017] A third aspect of the present invention provides a computer-readable storage medium having a program stored thereon, which, when executed by a processor, implements the steps of a sintering simulation homogenization method for ceramic lattice structures as described in the first aspect of the present invention.

[0018] The fourth aspect of the present invention provides an electronic device, including a memory, a processor, and a program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps in the sintering simulation homogenization method for ceramic lattice structures as described in the first aspect of the present invention.

[0019] The above one or more technical solutions have the following beneficial effects: (1) This invention uses ceramic sintering displacement as the core indicator for homogenization equivalence. By constructing a surrogate model of the sintering process based on a long short-term memory network and combining it with a differential evolution algorithm for global optimization, the equivalent constitutive parameters of the homogeneous structure are solved by inversion. This method uses a surrogate model to replace high-fidelity finite element simulation as a forward fast solver, which can achieve a high degree of consistency between the displacement time-varying curves of the homogeneous structure and the original lattice structure under the same sintering process with extremely low computational resource consumption, thereby significantly reducing the trial-and-error cost and computational resource consumption of large ceramic lattice structural components in the process development.

[0020] (2) This invention uses a long short-term memory network to extract temporal features and introduces an activation temperature gating mechanism to decouple the two physical effects of thermal expansion and sintering contraction. At the same time, it constructs a joint loss function containing a physical criterion penalty term (constraining that the settlement contraction in the direction of gravity is greater than the contraction in the horizontal direction), embedding physical constraints into the model training process. On this basis, a gradient-free differential evolution algorithm is used for global optimization, avoiding the numerical instability of gradient-based optimization methods in highly nonlinear parameter inversion, thereby achieving stable, efficient, and globally optimal inversion solutions for equivalent constitutive parameters.

[0021] Advantages of additional aspects of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0022] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an improper limitation of the invention.

[0023] Figure 1This is a flowchart of a sintering simulation homogenization method for ceramic lattice structural components according to Embodiment 1 of the present invention.

[0024] Figure 2 This is a schematic diagram of the network structure of the proxy model for the sintering process in Embodiment 1 of the present invention.

[0025] Figure 3 The actual lattice structure and homogeneous structure model in Embodiment 1 of the present invention are in X A comparative diagram of the shaft displacement results.

[0026] Figure 4 The actual lattice structure and homogeneous structure model in Embodiment 1 of the present invention are in Z A comparative diagram of the shaft displacement results.

[0027] Figure 5 This is a comparison diagram of the displacement field between the refined simulation and the homogenized equivalent sintering simulation of the ceramic lattice structure in Embodiment 1 of the present invention; wherein, Figure 5 In the figure, (a) represents the displacement field under the refined simulation of the ceramic lattice structure. Figure 5 In the figure, (b) represents the displacement field under the homogenized equivalent sintering simulation. Detailed Implementation

[0028] It should be noted that the following detailed descriptions are exemplary and intended to provide further illustration of the invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0029] It should be noted that the terminology used herein is for the purpose of describing particular implementations only and is not intended to limit the exemplary implementations of the present invention.

[0030] Where there is no conflict, the embodiments and features in the embodiments of the present invention can be combined with each other.

[0031] The overall approach of this invention is as follows: Since displacement is the macroscopic manifestation of structural deformation during sintering, directly affecting the dimensional accuracy and shape retention of components, and facilitating experimental measurement and simulation comparison, this invention provides a homogenization method for sintering simulation of ceramic lattice structures. This method uses displacement response as the core basis for homogenization equivalence. Specifically, compared to accurately simulating the deformation of complex lattice structures, an equivalent homogeneous structure is directly established. As long as the deformation curve of the homogeneous structure due to thermal shrinkage during simulation is consistent with that of the complex lattice structure, this homogeneous structure is used to replace the complex lattice for subsequent process optimization. By ensuring that the homogeneous structure and the original lattice structure have consistent displacement time-varying curves under the same sintering process, the inversion and equivalence of constitutive parameters can be achieved more intuitively and physically. Therefore, this method can transform the tedious sintering parameter calibration process into an efficient numerical optimization problem, thereby reducing reliance on physical experiments, reducing raw material loss and energy waste, and providing a low-cost, high-efficiency process simulation solution for the ceramic additive manufacturing industry, with significant economic benefits and social and environmental value.

[0032] Example 1 This embodiment discloses a sintering simulation homogenization method for ceramic lattice structural components.

[0033] like Figure 1 As shown, a sintering simulation homogenization method for ceramic lattice structural components includes: Step S1: Construct a lattice structure model and an equivalent homogeneous structure model, and obtain macroscopic displacement time series data of the ceramic lattice structure and homogeneous structure during the sintering simulation process; Step S2: Construct a sintering process proxy model based on a long short-term memory network to simulate the evolution of macroscopic displacement according to given constitutive parameters and temperature conditions. Step S3: Based on the joint loss function with physical constraints, pre-train the sintering process proxy model; Step S4: Perform global optimization based on differential evolution algorithm, and inversely solve the equivalent constitutive parameters of the homogeneous structure to make the displacement time-varying curves of the homogeneous structure and the ceramic lattice structure consistent under the same sintering process.

[0034] Based on the above process, this invention uses ceramic sintering displacement as the core indicator for homogenization equivalence, which can significantly reduce the trial-and-error costs and computational resource consumption in the process development of large ceramic lattice structures, thereby providing effective theoretical guidance for the manufacturing of ceramic additive manufacturing. To facilitate understanding of the technical solution of this invention, the specific implementation methods of this invention will be further explained and described below.

[0035] In step S1, a lattice structure model and an equivalent homogeneous structure model are constructed, and macroscopic displacement time series data of the ceramic lattice structure and homogeneous structure during the sintering simulation process are obtained.

[0036] The SOVS viscous flow model is used as the material constitutive model to construct a lattice structure model with fine porosity and a solid equivalent homogeneous structure model; wherein, the SOVS viscous flow model is expressed as: ; In the formula, Indicates inelastic strain rate; Indicates the fully dense shear modulus. Indicates the current temperature; Represents the normalized shear modulus. Represents the normalized bulk modulus. Indicates sintering stress; Represents relative density. Represents the deviatoric stress tensor; Represents the stress tensor traces, For Kronecker symbol.

[0037] Furthermore, shear modulus bulk modulus Sintering stress and fully dense shear modulus The specific expressions are as follows: ; ; ; ; in, , and These are the parameters to be fitted. It is the gas constant; Indicates local sintering stress. It is a natural constant.

[0038] The sintering process was simulated using finite element method (FEM) software. Sintering temperature loads were applied to the four surfaces of the model. A base plate was installed at the bottom of the model for heat conduction, and frictional constraints were set between the base plate and the bottom of the model. The entire model was then configured... Z Gravity in the negative direction of the axis; extraction X axis, Z The curve of axial displacement shrinkage over time is used as the equivalent standard experimental solution.

[0039] In the specific implementation process, a typical Triply Periodic Minimal Surface (TPMS) Gyroid cell was selected as the representative volume element, with a volume fraction of 30% and a cell size of 5mm × 5mm × 5mm. An equivalent simulation model for the lattice structure and the homogeneous structure was established. The material properties of the lattice structure were the mechanical and thermal properties of zirconia ceramic, while the material properties of the homogeneous structure were the equivalent mechanical and thermal properties of zirconia ceramic obtained through asymptotic homogenization equivalent calculation. The constitutive parameters (i.e., the parameters to be fitted) used in the lattice structure simulation were combined as follows: a 1=1、 b 1=2、 a 2 = 2 / 3 b 2=3、 c 2=1、 a 3=1、 b 3=2、 a 4 = 8.3 b 4=20. Within a reasonable range of constitutive parameters, uniform sampling is used to establish a combination of constitutive parameters for homogeneous structure simulation. The upper and lower limits of the constitutive parameters in this embodiment are shown in Table 1.

[0040] Table 1 Upper and lower limits of constitutive parameters

[0041] Based on the upper and lower limits of the constitutive parameters shown in Table 1, a dataset was constructed by using ABAQUS and CREEP subroutines for joint simulation to extract time-varying displacement data of the equivalent simulation model during a specific sintering curve process. It should also be noted that since the parameters to be fitted are all pre-exponential coefficients and exponents in the constitutive model, they are dimensionless constants.

[0042] In step S2, a sintering process proxy model based on a long short-term memory network (LSTM) is constructed to establish a fast solver that can replace complex finite element calculations, and to quickly simulate the evolution of macroscopic displacement based on given constitutive parameters and temperature conditions.

[0043] like Figure 2 The network structure shown is a surrogate model for the sintering process that combines deep learning and physical mechanisms. This surrogate model aims to address the deformation of materials during sintering due to thermal effects and densification. Its structure consists of three layers: First, an MLP encoder transmits static material parameters and time-varying temperature data. and relative density Feature fusion is performed; secondly, in the core temporal processing, LSTM units are used to capture the temporal evolution pattern, and a temperature-activated Sigmoid physical gate is used to control the two physical processes of "thermal expansion" and "sintering contraction"; finally, the displacement of the target lattice structure is combined with the feature fusion. relative density The time-varying data and the physical criteria are used to train the system, and the final output is an equivalent homogeneous structure. X shaft and Z Displacement-time curves on the axis. This architectural design enables the model to not only accurately fit experimental data but also strictly adhere to the thermodynamics and densification laws of the material, thereby effectively predicting and compensating for dimensional changes during sintering.

[0044] The input to the surrogate model for the sintering process consists of three parts: the macroscopic equivalent constitutive parameters to be inverted (i.e., the parameters to be fitted). , and The current time-step series being analyzed and the corresponding temperature loading sequence .

[0045] In the specific implementation process, firstly, a multilayer perceptron (MLP) is used as a parametric encoder to map the input static constitutive parameters into a high-dimensional feature vector for fusion with time-series information. Subsequently, the encoded feature vector, along with the time and temperature sequences, is input into the LSTM module. As a recurrent neural network with memory capabilities, LSTM, through its internal memory units and gating structure, can output a hidden feature vector at each time step. This vector integrates the input information at the current moment and the network's memory of the historical sintering process, effectively reflecting the cumulative effect of temperature changes on deformation behavior.

[0046] Step S2 can be implemented through the following process: 1) Input the code.

[0047] The nine macroscopic equivalent constitutive parameters to be inverted are input into a multilayer perceptron (MLP) and mapped to a 32-dimensional high-dimensional feature vector through two fully connected layers with SiLU activation functions. The encoded feature vectors can automatically extract the nonlinear coupling relationships between parameters and concatenate them with the process variables at each time step as the conditional input of the LSTM.

[0048] 2) Temporal feature extraction.

[0049] The encoded high-dimensional feature vector Time increment of process variables at the current moment Temperature increment Concatenate into a single input vector The data is fed into the LSTM unit step by step over time. The LSTM captures the temporal dependence of displacement and density evolution during sintering through its gating mechanism, and integrates parameter condition information to output a hidden feature vector containing the current state and historical memory. This hidden state encodes both thermodynamic and kinetic driving forces.

[0050] 3) Decoupling of physical state.

[0051] The hidden feature vectors output by the LSTM at each time step The anisotropic thermal expansion coefficient is obtained after the fully connected decoding branch is fed into it and activated by Softplus. , and irreversible sintering shrinkage rate , Thus, the hidden feature is explicitly decoupled into two physical channels: Thermal expansion channel: Calculates the instantaneous displacement increment caused by temperature changes. , ;in, The characteristic length is denoted as .

[0052] Sintering shrinkage channel: Calculate the irreversible displacement increment caused by sintering. , .

[0053] Simultaneously, based on the principle of mass conservation, the relative density increment of the current step is estimated from the shrinkage strain rate. and update density .

[0054] 4) Activate temperature gating.

[0055] Two physical effects exist simultaneously during sintering: reversible thermal expansion at low temperatures and irreversible sintering contraction at high temperatures. The hidden features output by the LSTM do not distinguish between these two effects; therefore, an activation temperature gating mechanism is needed to physically decompose and constrain them. Specifically, after separating the hidden features output by the LSTM into two parts, corresponding to the original activation values ​​of the thermal expansion channel and the sintering contraction channel respectively, a smoothing gating function related to the current temperature is introduced to calculate the weight that should be assigned to the sintering contraction channel at the current moment, i.e., the contraction stage gating value: ; in, The current temperature. The sintering activation temperature, Gating parameters for controlling smoothness.

[0056] The gating point during the thermal expansion phase is: ; The separated features are multiplied by the corresponding gate value for the thermal expansion stage. and the gate value during the contraction phase This yields the decoupled thermal expansion increment and the irreversible sintering shrinkage increment. When the temperature is much lower than the activation temperature... When the temperature approaches 0, the network primarily outputs a thermal expansion response; when the temperature exceeds the activation temperature, The network output gradually switches to be dominated by sintering shrinkage as it rapidly transitions to 1.

[0057] 5) Displacement accumulation: The thermal expansion increment and sintering shrinkage increment of each time step are superimposed to obtain the displacement increment of the current time step; the displacement increment is accumulated along the time axis to obtain the complete displacement-time curve.

[0058] In step S3, the sintering process proxy model is pre-trained based on a joint loss function with introduced physical constraints.

[0059] To avoid neural networks predicting results that violate physical laws, a joint loss function is constructed that includes displacement error and a penalty term for physical criteria. Specifically, it is expressed as: ; in, and Represent X Axial direction and Z Displacement loss in the axial direction, This represents a physical criterion penalty term, which is used to limit... Z The displacement in the direction is always greater than X The displacement in the direction of gravity is such that the settlement contraction in the direction of gravity is greater than the contraction in the horizontal direction.

[0060] This loss function not only limits the accuracy of displacement predictions by the neural network, but also mandates that the settlement contraction in the direction of gravity must be greater than the contraction in the horizontal direction. Through this multidimensional constraint, the surrogate model learns basic physical intuition during the pre-training phase.

[0061] In step S4, global optimization is performed based on the differential evolution algorithm to solve the equivalent constitutive parameters of the homogeneous structure, so that the displacement time-varying curves of the homogeneous structure and the ceramic lattice structure are consistent under the same sintering process.

[0062] The weights of all neurons in the pre-trained sintering process proxy model are frozen and used as a forward fast solver. Combined with a differential evolution global optimization algorithm, the population is initialized within the preset constitutive parameter boundaries and searched iteratively. The mean square error between the output of the sintering process proxy model and the displacement data of the ceramic lattice structure obtained in step S1 is used as the fitness function. Through crossover, mutation, and selection operations, iterative evolution is performed to search for the equivalent constitutive parameter combination that minimizes the fitness function. Specifically, the differential evolution inversion process is as follows: Within a space defined by upper and lower bounds of nine constitutive parameters, an initial population of size N is randomly generated, with each individual representing a set of candidate parameters. The frozen surrogate model is used as the forward solver, inputting the temperature and time series of the lattice structure, and quickly outputting the corresponding parameters for each candidate parameter. X , Z Orientation displacement and relative density prediction curves. The weighted mean square error between the predicted curve and the measured lattice structure curve in step S1 is used as the fitness function, where the density term is given a higher weight to enhance the fitting accuracy of the densification behavior.

[0063] The algorithm enters the iterative evolution phase: In each generation, a mutation vector is generated for each target individual in the population; the mutation vector is then mixed with the target individual using binomial crossover to generate a trial individual. The fitness of the trial individual is calculated; if it is better than the target individual, it is replaced and moved to the next generation; otherwise, the original individual is retained. This process of mutation, crossover, and greedy selection is repeated to drive the population towards a low-fitness region.

[0064] The termination condition is set to reaching the preset maximum number of generations, or the optimal fitness value showing no significant improvement over several consecutive generations. After the iteration ends, the individual with the best fitness in the population is output, which is the optimal equivalent homogenized constitutive parameter obtained through inversion.

[0065] This step can be understood as a mathematical optimization method simulating natural selection: first, candidate parameter combinations are randomly generated; then, the results are quickly calculated using a surrogate model; and finally, the deformation curve is compared to the lattice components in step S1. After iterative evolution, the parameter combinations that survive are the equivalent homogenization parameters capable of accurately simulating the sintering behavior of complex lattice structures. By minimizing the mean square error between the predicted target and the baseline target, the equivalent homogenization parameters that make the surrogate model output approximate the lattice structure data in step S1 are ultimately derived.

[0066] In this embodiment, the sintering process proxy model is used as the fitness evaluator. After multiple crossover, mutation, and selection iterations, the search is performed to find the model that... The optimal set of macroscopic equivalent parameters was minimized. A Gyroid-type 30% volume fraction lattice structure was established with a unit cell arrangement of 5*5*3 and a unit cell size of 5mm*5mm*5mm. The thickness of the upper and lower skins was 1mm. A corresponding homogeneous structural model was established, and sintering simulations were performed on the two models.

[0067] like Figure 3 , Figure 4 The figures shown are the actual lattice structure and the homogeneous structure model, respectively. X A comparative diagram of the axial displacement results, and the actual lattice structure and homogeneous structure model in... Z A comparative diagram of the shaft displacement results. The comparison results show that, in the displacement fitting process of a homogeneous structure, X The fitting accuracy of the axial displacement curve is 0.9998. Z The axial displacement curve fitting accuracy is 0.9997, demonstrating the high accuracy of the model in sintering scenarios.

[0068] Table 2 shows the equivalent homogenization constitutive parameter combination for homogeneous structure sintering simulation obtained by inversion using the sintering process proxy model and differential evolution algorithm. This parameter, combined with the homogeneous structure model, can be directly used to simulate the sintering process of lattice structures.

[0069] Table 2 Equivalent Homogenization Constitutive Parameter Combinations for Sintering Simulation of Homogeneous Structures

[0070] like Figure 5 The figure shows a comparison of the displacement field between refined simulation and homogenized equivalent sintering simulation of ceramic lattice structure. Figure 5 In the figure, (a) represents the displacement field under the refined simulation of the ceramic lattice structure. Figure 5 In the figure, (b) represents the displacement field under the homogenized equivalent sintering simulation. It can be seen from the displacement field cloud map that the displacement field distributions of the two models are highly overlapping and the numerical error is small.

[0071] Furthermore, the displacement comparison of the simulation results is shown in Table 3, that is, by extracting... Figure 4 Displacement simulation of mid-lattice structures and homogeneous structures X axis, Z Axis components yield two models. X shaft and Z The sintering shrinkage displacement in the axial direction was calculated. X shaft and Z The displacement error of the shaft is within 10%. Therefore, the parameters extracted by this invention can reconstruct the displacement change of the original lattice sintering process with higher accuracy.

[0072] Table 3 Comparison of sintering simulation displacement between lattice structure and equivalent homogeneous structure models

[0073] Example 2 This embodiment discloses a sintering simulation homogenization system for ceramic lattice structural components.

[0074] A sintering simulation homogenization system for ceramic lattice structural components includes: The structure building module is configured to: build a lattice structure model and an equivalent homogeneous structure model, and obtain macroscopic displacement time series data of the ceramic lattice structure and homogeneous structure during the sintering simulation process; The model building module is configured to: build a sintering process proxy model based on a long short-term memory network, which is used to simulate the evolution of macroscopic displacement based on given constitutive parameters and temperature conditions; The pre-training module is configured to pre-train the sintering process proxy model based on a joint loss function with introduced physical constraints. The simulation inversion module is configured to perform global optimization based on the differential evolution algorithm, and invert the solution of the equivalent constitutive parameters of the homogeneous structure so that the displacement time-varying curves of the homogeneous structure and the ceramic lattice structure are consistent under the same sintering process.

[0075] Example 3 The purpose of this embodiment is to provide a computer-readable storage medium.

[0076] A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps in a sintering simulation homogenization method for ceramic lattice structures as described in Embodiment 1 of this disclosure.

[0077] Example 4 The purpose of this embodiment is to provide an electronic device.

[0078] An electronic device includes a memory, a processor, and a program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps in a sintering simulation homogenization method for ceramic lattice structures as described in Embodiment 1 of this disclosure.

[0079] The steps and methods involved in the apparatuses of Embodiments 2, 3, and 4 above correspond to those in Embodiment 1. For specific implementation details, please refer to the relevant description section of Embodiment 1. The term "computer-readable storage medium" should be understood as a single medium or multiple media including one or more instruction sets; it should also be understood as including any medium capable of storing, encoding, or carrying an instruction set for execution by a processor and enabling the processor to perform any of the methods in this invention.

[0080] Those skilled in the art will understand that the modules or steps of the present invention described above can be implemented using general-purpose computer devices. Optionally, they can be implemented using computer-executable program code, thereby allowing them to be stored in a storage device for execution by a computer device, or they can be fabricated as separate integrated circuit modules, or multiple modules or steps can be fabricated as a single integrated circuit module. The present invention is not limited to any particular combination of hardware and software.

[0081] While the specific embodiments of the present invention have been described above in conjunction with the accompanying drawings, this is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art without creative effort based on the technical solutions of the present invention are still within the scope of protection of the present invention.

Claims

1. A method for homogenization in sintering simulation of ceramic lattice structures, characterized in that, include: A lattice structure model and an equivalent homogeneous structure model are constructed to obtain macroscopic displacement time-series data of the ceramic lattice structure and homogeneous structure during the sintering simulation process. Specifically, the SOVS viscous flow model is used as the material constitutive model to construct a lattice structure model with fine porosity and a solid equivalent homogeneous structure model. The SOVS viscous flow model is expressed as follows: ； In the formula, Indicates inelastic strain rate; Indicates the fully dense shear modulus. Indicates the current temperature; Represents the normalized shear modulus. Represents the normalized bulk modulus. Indicates sintering stress; Represents relative density. Represents the deviatoric stress tensor; Represents the stress tensor traces, The symbol for Kronecker; A sintering process proxy model based on a long short-term memory network is constructed to simulate the evolution of macroscopic displacement according to given constitutive parameters and temperature conditions. The construction of the sintering process proxy model is achieved through sequential input encoding, temporal feature extraction, physical state decoupling, activation of temperature gating, and displacement accumulation. The activation of temperature gating is achieved through a preset temperature gating mechanism, and the calculation formulas for the gating values ​​in the contraction and thermal expansion stages are expressed as follows: ； ； in, The current temperature. The sintering activation temperature, These are the gating parameters used to control smoothness; and These represent the gate values ​​for the contraction and thermal expansion phases, respectively. The sintering process proxy model is pre-trained based on a joint loss function with physical constraints. Global optimization is performed based on the differential evolution algorithm, and the equivalent constitutive parameters of the homogeneous structure are solved by inversion, so that the displacement time-varying curves of the homogeneous structure and the ceramic lattice structure are consistent under the same sintering process.

2. The sintering simulation homogenization method for ceramic lattice structural components as described in claim 1, characterized in that, The joint loss function Represented as: ； in, and Represent X Axial direction and Z Displacement loss in the axial direction, This represents a physical criterion penalty term, which is used to limit... Z The displacement in the direction is always greater than X The displacement in the direction of gravity is such that the settlement contraction in the direction of gravity is greater than the contraction in the horizontal direction.

3. The sintering simulation homogenization method for ceramic lattice structural components as described in claim 1, characterized in that, Global optimization based on the differential evolution algorithm is used to inversely solve for the equivalent constitutive parameters of the homogeneous structure, including: The weights of the pre-trained sintering process surrogate model are frozen and used as a forward fast solver. Combined with the differential evolution global optimization algorithm, the population is initialized within the preset constitutive parameter boundary. The mean square error between the output of the sintering process surrogate model and the displacement data of the obtained ceramic lattice structure is used as the fitness function. Through crossover, mutation and selection operations, iterative evolution is carried out to search for the equivalent constitutive parameter combination that minimizes the fitness function.

4. The sintering simulation homogenization method for ceramic lattice structural components as described in claim 1, characterized in that, The ceramic lattice structure is a three-period minimal surface structure, which is specifically represented as a Gyroid-type cell structure.

5. A sintering simulation homogenization system for ceramic lattice structures, employing the sintering simulation homogenization method as described in any one of claims 1-4, characterized in that, include: The structure building module is configured to: build a lattice structure model and an equivalent homogeneous structure model, and obtain macroscopic displacement time series data of the ceramic lattice structure and homogeneous structure during the sintering simulation process; The model building module is configured to: build a sintering process proxy model based on a long short-term memory network, which is used to simulate the evolution of macroscopic displacement based on given constitutive parameters and temperature conditions; The pre-training module is configured to pre-train the sintering process proxy model based on a joint loss function with introduced physical constraints. The simulation inversion module is configured to perform global optimization based on the differential evolution algorithm, and invert the solution of the equivalent constitutive parameters of the homogeneous structure so that the displacement time-varying curves of the homogeneous structure and the ceramic lattice structure are consistent under the same sintering process.

6. A computer-readable storage medium having a program stored thereon, characterized in that, When executed by the processor, the program implements the steps in the sintering simulation homogenization method for ceramic lattice structures as described in any one of claims 1-4.

7. An electronic device, comprising a memory, a processor, and a program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the steps in the sintering simulation homogenization method for ceramic lattice structures as described in any one of claims 1-4.

Images