A machine learning-assisted melt infiltration modification method for preparing carbon-carbon composite materials

Abstract

This invention discloses a machine learning-assisted melt infiltration modification method for preparing carbon-carbon composite materials, comprising: constructing several reactive melt infiltration candidate compositions; using machine learning methods and based on melt infiltration element data of each candidate composition to predict the linear ablation rate and mass ablation rate of each candidate composition; using the predicted values ​​of linear ablation rate and mass ablation rate as dual-objective optimization variables to extract the Pareto front in the constructed dual-objective optimization space; calculating the distance from each reactive melt infiltration candidate composition on the Pareto front to a preset ideal point based on the linear ablation rate and mass ablation rate, and selecting the optimal reactive melt infiltration composition based on the distance; preparing a melt infiltrator according to the melt infiltration element composition information corresponding to the optimal reactive melt infiltration composition, and performing reactive melt infiltration treatment on the carbon-carbon composite matrix according to the corresponding preparation process parameters and ablation environment parameters. This invention can improve the service performance of the prepared carbon-carbon composite material under extreme thermo-oxidative environments.

Description

Technical Field

[0001] This invention relates to the field of computational materials science and technology, specifically to a method for preparing carbon-carbon composite materials with machine learning-assisted melt infiltration modification. Background Technology

[0002] Carbon / carbon composites possess low density, high specific strength, high specific modulus, excellent thermal shock resistance, and high-temperature mechanical stability, making them valuable for applications in aerospace thermal protection, solid rocket engine nozzles, aircraft nose cones, brake materials, and other components used in extreme thermal environments. However, carbon / carbon composites are prone to rapid oxidation and ablation in high-temperature, oxygen-containing environments, leading to mass loss, dimensional degradation, and reduced load-bearing capacity, thus limiting their service reliability under prolonged, high-heat-flux, and high-oxygen-pressure conditions.

[0003] To improve the oxidation and ablation resistance of carbon / carbon composites, reactive melt infiltration modification is a commonly used and effective ceramicization modification method. This method typically uses high-temperature ceramic-forming elements such as Si, Zr, Hf, V, Mo, Ta, and Ti as infiltration components. These elements penetrate the open pores, fiber bundle gaps, and matrix cracks of the carbon / carbon composite under high-temperature conditions, reacting with the carbon matrix or pyrolytic carbon to generate ceramic phases such as SiC, ZrC, HfC, and VC. These ceramic phases can, on the one hand, increase the material's density and reduce the channels for oxygen diffusion into the carbon matrix; on the other hand, under ablation conditions, they can form a protective layer of SiO2, ZrO2, HfO2, V-based oxides, or composite oxides. This improves the material's ablation resistance through mechanisms such as liquid-phase pore sealing, solid-phase framework support, and oxygen diffusion barrier.

[0004] The Si-Zr-Hf-V multi-element reactive melt infiltration system possesses several potential advantages. The Si component can form SiC, and after oxidation, generate SiO2 or a Si-containing glass phase with viscous flow sealing capabilities. The Zr and Hf components can form high-melting-point ZrC, HfC, and corresponding ZrO2 and HfO2 solid-phase oxide frameworks, providing high-temperature stability. The V component and its oxidation products can regulate the fluidity of the oxide layer, sintering behavior, and the formation process of composite oxides. Therefore, the Si-Zr-Hf-V system is expected to achieve excellent anti-oxidation and ablation performance through the synergistic effect of liquid-phase protection and solid-phase protection. However, this quaternary system has a large compositional space, and different element ratios have complex effects on melt infiltration densification, ceramic phase formation, oxide product composition, protective layer structure, linear ablation rate, and mass ablation rate. Relying solely on traditional experimental trial-and-error methods is insufficient to efficiently determine the optimal composition.

[0005] In existing technologies, the following problems are commonly encountered in the composition design and ablation performance optimization of reactive melt-infiltrated modified carbon / carbon composites.

[0006] First, traditional composition optimization mainly relies on empirical design and point-by-point experimental verification, which is inefficient. For the Si-Zr-Hf-V quaternary system, the types of elements are fixed but the combinations are numerous. If a full combination experiment screening is adopted, not only will the experimental cost be high and the cycle be long, but it will also be difficult to systematically reveal the synergistic effect between different elements.

[0007] Second, existing machine learning methods mostly predict material properties directly using the original elemental ratios, density, porosity, and process parameters, lacking an expression of the physical processes of reactive melting and infiltration. Reactive melting and infiltration involves continuous processes such as liquid phase penetration, interface wetting, carbonization reaction, pore shrinkage, ceramic phase filling, and the formation of the post-melting and infiltrated state. If the ablation rate is predicted directly from the original composition and process parameters, key intermediate physical factors such as post-melting and infiltrated density, pore filling efficiency, carbonization reaction driving force, and transport-reaction coupling are easily overlooked, leading to insufficient model interpretability.

[0008] Third, ablation performance is not determined by a single indicator. Linear ablation rate primarily reflects the degree of degradation in the thickness direction or surface morphology of the material, while mass ablation rate primarily reflects the overall mass loss behavior of the material. Both are influenced by factors such as densification degree, oxidation product volatilization, liquid phase loss, solid-phase skeleton exfoliation, and airflow erosion, and may exhibit inconsistent optimization trends. Therefore, optimizing solely based on linear ablation rate or a single ablation indicator is insufficient to comprehensively evaluate the ablation resistance of materials; both linear ablation rate and mass ablation rate need to be considered simultaneously, and a dual-objective optimization screening process should be implemented.

[0009] Fourth, under conditions of small sample material data, directly using high-dimensional original variables for modeling is prone to overfitting and insufficient generalization ability. The preparation and ablation experiments of carbon / carbon composite materials are costly and time-consuming, and the experimental samples are usually limited. If there is a lack of physically inspired features to reduce the dimensionality of the data and enhance the mechanism, the machine learning model is prone to learning random statistical correlations and is difficult to stably extrapolate to the unknown Si-Zr-Hf-V quaternary composition space.

[0010] Fifth, although the carbide ceramic phase formed after reactive infiltration can improve ablation resistance, it may still encounter problems such as oxygen diffusion, oxide layer cracking, pore penetration, and protective layer failure under extreme high-temperature oxidation environments. Nitriding treatment, which at least partially converts ZrC, HfC, and / or VC into carbonitrides or nitrides, is expected to improve the ceramic phase's ability to impede oxygen atom diffusion and further enhance the material's resistance to oxidation and ablation. However, current research on the hot isostatic pressing nitriding treatment and oxygen barrier mechanism of Si-Zr-Hf-V modified carbon / carbon composites after reactive infiltration is still insufficient, especially lacking a method to explain the mechanism of oxygen atom diffusion barrier changes using first-principles calculations. Summary of the Invention

[0011] This invention provides a machine learning-assisted melt infiltration modification method for preparing carbon-carbon composite materials, which improves the service performance of the prepared carbon-carbon composite materials under extreme thermo-oxidative environments.

[0012] To achieve the above technical objectives, the present invention adopts the following technical solution: A machine learning-assisted melt infiltration modification method for preparing carbon-carbon composite materials includes: Step 1: Construct several reactive melting candidate compositions. Using machine learning methods, based on the melting element composition information, pore structure parameters, preparation process parameters, preform density parameters, and ablation environment parameters of each candidate composition, predict the linear ablation rate and mass ablation rate of each candidate composition. Step 2: Using the predicted linear ablation rate and the predicted mass ablation rate as bi-objective optimization variables, extract the non-dominated solution set in the constructed bi-objective optimization space and extract the Pareto front. Step 3: Calculate the distance from each reactive infiltration candidate composition to the preset ideal point on the Pareto front based on the linear ablation rate and the mass ablation rate, and screen the optimal reactive infiltration composition based on the distance. Step 4: Prepare a melting agent according to the melting element composition information corresponding to the optimal reaction melting composition, and perform reaction melting treatment on the carbon-carbon composite matrix according to the preparation process parameters and ablation environment parameters corresponding to the optimal reaction melting composition.

[0013] Furthermore, all of the candidate compositions for reactive infiltration are selected from the Si-Zr-Hf-V infiltration system.

[0014] Furthermore, after obtaining the Si-Zr-Hf-V reactive melt infiltration modified carbon-carbon composite material by performing reactive melt infiltration treatment in step 4, it is further subjected to hot isostatic pressing nitriding treatment to obtain the Si-Zr-Hf-VN modified carbon-carbon composite material.

[0015] Furthermore, when extracting the non-dominated solution set in step 2, if the predicted linear ablation rate and the predicted mass ablation rate corresponding to a certain reactive infiltration candidate composition are both no greater than the other candidate composition in terms of both linear ablation rate and mass ablation rate, and at least one of the targets is less than the other candidate composition, then the former is considered to dominate the latter; all candidate compositions that are not dominated by other candidate compositions constitute the Pareto front.

[0016] Furthermore, the preset ideal point in step 3 is: The distance calculated based on the linear ablation rate and the mass ablation rate is expressed as follows: ; in, The distance from the candidate composition for reactive melt infiltration to the preset ideal point. and The linear ablation rate and mass ablation rate of the reactive infiltration candidate composition are respectively.

[0017] Furthermore, the reactive infiltration treatment includes: The prepared infiltration agent is mixed and then placed in contact with the carbon-carbon composite matrix; The temperature is raised to 1400–2000℃ under vacuum, argon, helium, or a protective atmosphere and held for 0.5–4 hours, allowing the infiltrator to enter the open pores, fiber bundle gaps, and matrix microcracks of the carbon-carbon composite material and react with the carbon matrix or pyrolytic carbon to generate one or more ceramic phases of SiC, ZrC, HfC, and VC.

[0018] Furthermore, the hot isostatic nitriding treatment is carried out in a nitrogen, ammonia, nitrogen-argon mixture, nitrogen-hydrogen mixture, or nitrogen-containing atmosphere, with a nitriding temperature of 1600–2000℃, a nitriding pressure of 150–200MPa, and a holding time of 0.5–8h.

[0019] Furthermore, machine learning methods are used to predict the linear ablation rate and mass ablation rate corresponding to each candidate component. Specifically: Based on the elemental composition information of the melt infiltration, pore structure parameters, preparation process parameters, and preform density parameters, a physical-inspired feature for characterizing the density after melt infiltration is constructed. Based on the predicted density after melting and infiltration, as well as information on the elemental composition of the melt infiltration, pore structure parameters, preparation process parameters, preform density parameters, and ablation environment parameters, physical heuristic features for characterizing ablation protection behavior are constructed. A density prediction model after melting and infiltration is constructed and trained using a training set. The input of the model is the physical heuristic features used to characterize the density after melting and infiltration, and the output is the density after melting and infiltration. A linear ablation rate prediction model and a mass ablation rate prediction model were constructed and trained using a training set. The inputs to the linear ablation rate prediction model and the mass ablation rate prediction model are both physical heuristic features used to characterize ablation protection behavior, and the outputs are linear ablation rate and mass ablation rate, respectively. Using a trained post-melting density prediction model, the corresponding post-melting density is predicted based on the physical heuristic features of each reactive melting candidate composition used to characterize ablation protection behavior. Then, using the trained linear ablation rate prediction model and mass ablation rate prediction model, the corresponding linear ablation rate and mass ablation rate are predicted respectively based on the physical heuristic features of each reactive infiltration candidate composition used to characterize the ablation protection behavior.

[0020] Furthermore, the physical heuristic features used to characterize ablation protection behavior include one or more of the following: pyrolytic carbon deposition densification features, carbonization reaction enthalpy weighted features, transport-reaction coupling features, and penetration driving force features; The pyrolysis carbon deposition densification characteristics are characterized by calculating the following features: pyrolysis carbon deposition increment characteristics. Relative densification characteristics of pyrolytic carbon Pyrolysis carbon filling efficiency characteristics : ; ; ; In the formula, Density before deposition Density after deposition Open area ratio; The enthalpy-weighted characteristic of the carbonization reaction is characterized by calculating the following feature: the total component carbonization reaction enthalpy weighted value. Normalized total component carbonization reaction enthalpy characteristics Enthalpy-weighted values ​​of carbonization reactions of non-Cu active carbide elements Average enthalpy characteristics of non-Cu active carbide elements Ratio of active carbide forming elements : ; ; ; ; ; In the formula, For infiltration elements atomic ratio fraction, For infiltration elements Enthalpy of carbonization reaction; The element is Cu; The transport-reaction coupling characteristic is a composite characteristic calculated from porosity, carbonization volume expansion, average atomic radius of active elements, melting point of carbides corresponding to active elements, and preparation temperature. ; In the formula, It is a composite feature; The weighted average of the carbonization volume expansion of multi-element modification for high-temperature ceramics. This is the weighted average of the average atomic radii of non-Cu reactive elements. This is the weighted average of the melting points of the carbides corresponding to non-Cu reactive elements. The preparation temperature; The permeation driving force characteristic is denoted as The average cohesive energy, average electronegativity difference, average atomic weight, and average melting point were calculated based on the weighted average atomic ratios of the infiltrating elements. ; In the formula, The average cohesive energy is obtained by weighting the atomic ratio of the infiltrating elements. The average electronegativity difference of elements relative to carbon atoms, obtained by weighting according to the atomic ratio. The average atomic weight is obtained by weighting the atomic ratios of the elements. The average melting point is obtained by weighting the atomic ratios of the elements; where the average cohesive energy is... Poor average electronegativity Average atomic weight Average melting point They are uniformly denoted as equivalent physical properties. The weighted average of the physical properties of each element is obtained based on the atomic ratio of the infiltrating elements. ; In the formula, For infiltration elements atomic ratio, For infiltration elements Property characteristics to be determined The corresponding physical property parameters.

[0021] Furthermore, the physical heuristic features used to characterize ablation protection behavior include one or more of the following: basic protection features, liquid phase protection features, solid phase protection features, and liquid-solid equilibrium features; The basic protective features are used to characterize the coupling effect of the material's initial skeleton density, the degree of melt infiltration and densification retention, and the effectiveness of pore filling; The liquid phase protection feature is used to characterize the formation, flow, spreading, and blocking ability of liquid or viscous oxidation products under high-temperature ablation conditions; The solid-phase protection features are used to characterize the volume reserve, thermal stability, and resistance to high-temperature instability of the high-melting-point solid-phase oxide framework. The liquid-solid equilibrium characteristics are used to characterize the relative dominance between liquid-phase protection mechanisms and solid-phase protection mechanisms.

[0022] Furthermore, the formula for calculating the basic protection feature is: ; ; ; ; in, Represents basic protection features. Indicates density before deposition. For melt penetration retention rate, To improve melt infiltration filling efficiency, This is the predicted density value after melt infiltration. Density after CVI The equivalent mixing density of the melt-infiltrated phase, For infiltration elements mass fraction, For element infiltration metal density, Indicates the open area ratio;

[0023] The formula for calculating the liquid phase protection feature is: ; ; ; ; ; in, Represents liquid phase protection characteristics, This represents the total volume of the liquid-phase oxidation products. The average optical alkalinity in the liquid phase. The average ion field strength in the liquid phase; For the normalized characteristics of heat load, For the original heat load, This is the average of the heat load samples. Oxygen flow rate, For gas flow rate, For ablation time, This refers to the distance between the spray gun and the spray gun. This represents the normalized characteristic of the oxidative load. For the original oxidative load, This represents the mean of the oxidative load samples. The formula for calculating the solid-phase protection characteristics is: ; in, Represents solid-phase protection characteristics. This represents the total volume of the solid-phase oxidation products. This is an indicator of solid-phase stability. The formula for calculating the liquid-solid equilibrium characteristics is: ; in, Represents liquid-solid equilibrium characteristics. To prevent extremely small positive numbers with a denominator of zero.

[0024] To address the problems of existing reactive melt-infiltration modified carbon / carbon composite material composition design relying mainly on empirical trial and error, insufficient expression of ablation performance prediction mechanisms, and difficulty in simultaneously optimizing linear ablation rate and mass ablation rate, this invention provides a machine learning-assisted melt-infiltration modified carbon-carbon composite material preparation method, which has the following advantages compared with existing technologies:

[0025] 1. This invention uses machine learning methods to simultaneously predict linear ablation rate and mass ablation rate, taking into account both linear ablation damage and mass loss behavior. Furthermore, by screening reactive melt infiltration compositions with low linear ablation rate and mass ablation rate through Pareto front screening, it can reduce a large number of empirical trial and error experiments and improve the service performance of the prepared carbon-carbon composite material in extreme thermo-oxidative environments.

[0026] 2. This invention simultaneously establishes a density prediction model, a linear ablation rate prediction model, and a mass ablation rate prediction model after melting and infiltration, which can take into account the degree of material densification, linear ablation damage, and mass loss behavior.

[0027] 3. This invention introduces carbonitride or nitride phases through hot isostatic pressing nitriding treatment, and explains the mechanism of oxygen diffusion barrier elevation through first-principles calculations, which is beneficial to further improve the oxidation and ablation resistance of Si-Zr-Hf-V modified carbon / carbon composite materials.

[0028] 4. This invention divides the ablation performance optimization process into multiple levels: "whitelist data collection - physical heuristic feature construction - density prediction after melting and infiltration - linear / mass ablation rate prediction - quaternary space optimization screening - reactive melting and infiltration preparation - hot isostatic pressing nitriding - DFT mechanism verification", forming a closed-loop design route from data prediction to material preparation.

[0029] 5. This invention improves the physical interpretability and small-sample generalization ability of machine learning models by using pyrolysis carbon deposition densification features, carbonization reaction enthalpy weighting features, transport-reaction coupling features, permeation driving force features, and liquid / solid phase protection features. Attached Figure Description

[0030] Figure 1 This is a flowchart of the method for preparing carbon-carbon composite materials with machine learning-assisted melt infiltration modification as described in the embodiments of this application. Detailed Implementation

[0031] The embodiments of the present invention will be described in detail below. These embodiments are based on the technical solutions of the present invention and provide detailed implementation methods and specific operation processes to further explain the technical solutions of the present invention.

[0032] Example 1

[0033] This embodiment provides a machine learning-assisted melt infiltration modification method for preparing carbon-carbon composite materials, such as... Figure 1 As shown, it includes the following steps.

[0034] Step 1: Construct several reactive melting candidate compositions. Using machine learning methods, based on the melting element composition information, pore structure parameters, preparation process parameters, preform density parameters, and ablation environment parameters corresponding to each candidate composition, predict the linear ablation rate and mass ablation rate of each candidate composition.

[0035] In this embodiment, the candidate compositions for reactive melting are all selected from the Si-Zr-Hf-V melting infiltration system.

[0036] First, experimental data on carbon / carbon composites modified using the Si-Zr-Hf-V whitelist melt infiltration system were collected. Each set of data included melt infiltration elemental composition information, pore structure parameters, preparation process parameters, preform density parameters, and ablation environment parameters.

[0037] (1) Information on the composition of the infiltrating elements. The information on the composition of the infiltrating elements includes the types of infiltrating elements, their atomic ratios, and physical property parameters. In this embodiment, the elements in the Si-Zr-Hf-V infiltrating system include one or more of Zr, Hf, Si, and V. The physical property parameters include metal melting point, electronegativity, atomic weight, atomic radius, cohesive energy, carbide melting point, and carbide volume expansion.

[0038] (2) Pore structure parameters.

[0039] The pore structure parameters include the porosity. In this embodiment, the porosity is converted to a decimal form; if the input value is greater than 1, it is converted by dividing the percentage by 100.

[0040] (3) Preparation process parameters.

[0041] The preparation process parameters include the reaction melting and infiltration preparation temperature, and in other embodiments, the heating rate, holding time, and pressure may also be included.

[0042] (4) Precast density parameters.

[0043] The density parameters of the preform include pre-deposition density, post-deposition density, post-CVI density, and post-melting density.

[0044] (5) The ablation environment parameters include oxygen flow rate, gas flow rate, ablation time and spray gun distance.

[0045] The experimental data is then preprocessed, including unit standardization, porosity conversion, elemental composition normalization, missing value handling, and outlier removal. If the porosity is input as a percentage, it is converted to a decimal.

[0046] Subsequently, based on the elemental composition information of the melt infiltration, pore structure parameters, preparation process parameters, and density parameters before and after deposition, physical heuristic features for characterizing the density after melt infiltration are constructed. The physical heuristic features for characterizing ablation protection behavior described in this embodiment include one or more of the following: pyrolytic carbon deposition densification features, carbonization reaction enthalpy weighted features, transport-reaction coupling features, infiltration driving force features, and system-level equivalent physical property features.

[0047] The pyrolytic carbon deposition densification characteristics include pyrolytic carbon deposition increment characteristics. Relative densification characteristics of pyrolytic carbon Pyrolysis carbon filling efficiency characteristics These are used to characterize the absolute density, relative density, and filling efficiency per unit open space of the pyrolysis carbon deposition process, respectively, and are calculated based on the pre-deposition density, post-deposition density, and open area ratio. ; ; ; In the formula, Density before deposition Density after deposition This refers to the open area ratio.

[0048] The enthalpy-weighted characteristic of carbonization reaction includes the total component carbonization reaction enthalpy weighted value. Normalized total component carbonization reaction enthalpy characteristics Enthalpy-weighted values ​​of carbonization reactions of non-Cu active carbide elements Average enthalpy characteristics of non-Cu active carbide elements Ratio of active carbide forming elements This is used to characterize the overall carbonization reaction driving force of the system and the contribution of active carbonizing elements to the melt infiltration reaction. Each characteristic is obtained by weighting the melt infiltration elemental composition information and the carbonization reaction enthalpy of each element: ; ; ; ; ; In the formula, For infiltration elements atomic ratio fraction, For infiltration elements Enthalpy of carbonization reaction; The element is Cu. For the Si-Zr-Hf-V infiltration system in this embodiment, Si, Zr, Hf, and V can all be used as active carbide elements in the above calculations.

[0049] The transport-reaction coupling characteristic is a composite feature calculated from porosity, carbonization volume expansion, average atomic radius of active elements, melting point of the corresponding carbide of active elements, and preparation temperature. It is used to characterize the competitive relationship between liquid phase infiltration, reaction generation, and pore shrinkage or blockage during the melting and infiltration process, reflecting the nonlinear mechanism of transport and reaction working together during density formation after melting and infiltration. The calculation formula is: ; In the formula, It is a composite feature; This is the weighted average of the carburization volume expansion of the Si-Zr-Hf-V melt infiltration system. This is the weighted average of the average atomic radii of non-Cu reactive elements. This is the weighted average of the melting points of the carbides corresponding to non-Cu reactive elements. The preparation temperature.

[0050] Besides the pore structure and reaction product characteristics, the densification process after melt infiltration is also affected by the physicochemical properties of the infiltrating components themselves. To comprehensively characterize the effective infiltration capacity of multi-component components at a given preparation temperature, this embodiment calculates the infiltration driving force characteristics based on the average cohesive energy, average electronegativity difference, average atomic weight, and average melting point obtained by weighting the atomic ratios of the infiltrating elements. : ; In the formula, The average cohesive energy is obtained by weighting the atomic ratio of the infiltrating elements. The average electronegativity difference of elements relative to carbon atoms, obtained by weighting according to the atomic ratio. The average atomic weight is obtained by weighting the atomic ratios of the elements. The average melting point is obtained by weighting the atomic ratios of elements. The average cohesive energy... Poor average electronegativity Average atomic weight Average melting point They are uniformly denoted as equivalent physical properties. The weighted average of the physical properties of each element is obtained based on the atomic ratio of the infiltrating elements. ; In the formula, For infiltration elements atomic ratio, For infiltration elements Property characteristics to be determined The corresponding physical property parameters.

[0051] Since a greater electronegativity difference between the infiltrating element and carbon generally implies stronger interfacial chemistry and a greater tendency to react, while higher cohesive energy characterizes a more significant interatomic binding energy scale; conversely, a larger average atomic weight and a higher average melting point raise the threshold for system flow, wetting, and reaction. Therefore, this embodiment uses... The introduction of a thermal activation inhibition effect reflects the decrease in penetration propulsion capability when the process temperature is below the characteristic melting point of the component. Therefore, this penetration driving force characteristic... This essentially characterizes the competitive relationship between "interfacial chemical driving" and "thermal property threshold inhibition" in a multi-component melt infiltration system, and can serve as an important physical inspiration feature affecting melt infiltration densification behavior.

[0052] Using the aforementioned pyrolysis carbon deposition densification characteristics, carbonization reaction enthalpy weighted characteristics, transport-reaction coupling characteristics, infiltration driving force characteristics, and system-level equivalent physical property characteristics as inputs, and the measured post-melt-infiltration density as output, a post-melt-infiltration density prediction model is trained. The post-melt-infiltration density prediction model employs a random forest regression model, an XGBoost model, an extreme random tree regression model, or an ensemble model thereof, and the model parameters are determined through cross-validation and hyperparameter optimization.

[0053] After obtaining the density prediction model after melting and infiltration, physical heuristic features of ablation protection behavior are further constructed. The physical heuristic features used to characterize ablation protection behavior in this embodiment include one or more of the following: basic protection features, liquid phase protection features, solid phase protection features, and liquid-solid equilibrium features.

[0054] The aforementioned basic protective features characterize the coupling effect of the material's initial skeleton density, the degree of retention of melt infiltration densification, and the effectiveness of pore filling. These features are calculated based on the density after melt infiltration, the density after CVI, the density before deposition, the porosity, and the equivalent mixed density of the melt-infiltrated phase. ; ; ; ; in, Represents basic protection features. Indicates density before deposition; For melt penetration retention rate, To improve melt infiltration filling efficiency, This is the predicted density value after melt infiltration. Density after CVI The equivalent mixing density of the melt-infiltrated phase, For infiltration elements mass fraction, For infiltration elements metal density, This indicates the open area ratio.

[0055] The liquid phase protection characteristics are used to characterize the formation, flow, spreading, and blocking ability of liquid or viscous oxidation products under high-temperature ablation conditions. These characteristics are calculated based on the volume of the liquid phase oxidation products, the average optical alkalinity of the liquid phase, the average ionic field strength of the liquid phase, and the environmental thermal and oxygen load. ; ; ; ; ; in, Represents liquid phase protection characteristics, This represents the total volume of the liquid-phase oxidation products. The average optical alkalinity in the liquid phase. The average ion field strength in the liquid phase; For the normalized characteristics of environmental heat load, As the original heat load of the environment, This represents the average environmental heat load sample. Oxygen flow rate, For gas flow rate, For ablation time, This refers to the distance between the spray gun and the spray gun. This represents the normalized characteristics of environmental oxidative load. As the original oxidative load of the environment, This represents the sample mean of the environmental oxidative load.

[0056] The solid-phase protection characteristics are used to characterize the volume reserve, thermal stability, and resistance to high-temperature instability of the high-melting-point solid-phase oxide framework. They are calculated based on the volume of solid-phase oxidation products, solid-phase stability indices, and environmental thermal and oxygen loads. ; in, Represents solid-phase protection characteristics. This represents the total volume of the solid-phase oxidation products. This is an indicator of solid-phase stability.

[0057] The liquid-solid equilibrium characteristics are used to characterize the relative dominance between the liquid-phase protection mechanism and the solid-phase protection mechanism, and are calculated based on the volumes of liquid-phase oxidation products and solid-phase oxidation products. ; in, Represents liquid-solid equilibrium characteristics. To prevent extremely small positive numbers with a denominator of zero.

[0058] Using the aforementioned physical heuristic features for characterizing ablation protection behavior as input, and linear ablation rate and mass ablation rate as output, respectively, the linear ablation rate prediction model and the mass ablation rate prediction model are trained.

[0059] The Si-Zr-Hf-V melt infiltration system in this embodiment includes one or more of Si, Si-Zr, Si-Hf, Si-V, Zr-Hf, Zr-V, Hf-V, Si-Zr-Hf, Si-Zr-V, Si-Hf-V, Zr-Hf-V, and Si-Zr-Hf-V. The candidate composition space for constructing the Si-Zr-Hf-V melt infiltration system in this embodiment is: ; in, These represent the elements added during melt infiltration. The atomic ratio is set within the range of this embodiment: It ranges from 20% to 70%. It ranges from 5% to 40%. It ranges from 5% to 40%. The composition ranges from 1% to 30%, with a step size of 1%. For each candidate composition, a densification behavior physical heuristic feature is constructed in the same manner, and input into the post-melting density prediction model to obtain the post-melting density prediction value. Then, based on the post-melting density prediction value, an ablation protection behavior physical heuristic feature is constructed, and input into the linear ablation rate prediction model and the mass ablation rate prediction model respectively to obtain the linear ablation rate prediction value. and predicted mass ablation rate .

[0060] Step 2: Using the predicted linear ablation rate and the predicted mass ablation rate as bi-objective optimization variables, extract the non-dominated solution set in the constructed bi-objective optimization space and extract the Pareto front.

[0061] When extracting the non-dominated solution set, if the predicted linear ablation rate and the predicted mass ablation rate of a certain reaction infiltration candidate composition are both no greater than the other candidate composition in terms of both linear ablation rate and mass ablation rate, and at least one of the targets is less than the other candidate composition, then the former is considered to dominate the latter; all candidate compositions that are not dominated by other candidate compositions constitute the Pareto front.

[0062] Step 3: Calculate the distance from each reactive infiltration candidate composition to the preset ideal point on the Pareto front based on the linear ablation rate and the mass ablation rate, and select the optimal reactive infiltration composition based on the distance.

[0063] The ideal point preset in this embodiment is: The distance calculated based on the linear ablation rate and the mass ablation rate is expressed as follows: ; in, The distance from the candidate composition for reactive melt infiltration to the preset ideal point. and The linear ablation rate and mass ablation rate of the reactive infiltration candidate composition are respectively.

[0064] According to distance Sort by size from smallest to largest, select one or more candidate compositions with the smallest distance as preferred Si-Zr-Hf-V quaternary melt infiltration compositions for subsequent reaction melt infiltration and hot isostatic pressing nitriding preparation.

[0065] Step 4: Prepare a melting agent according to the melting element composition information corresponding to the optimal reaction melting composition, and perform reaction melting treatment on the carbon-carbon composite matrix according to the preparation process parameters and ablation environment parameters corresponding to the optimal reaction melting composition.

[0066] The reaction infiltration treatment includes: mixing the prepared infiltration agent and placing it in contact with the carbon-carbon composite matrix; then heating it to 1400-2000℃ under vacuum, argon, helium or protective atmosphere and holding it at that temperature for 0.5-4h, so that the infiltration agent enters the open pores, fiber bundle gaps and matrix microcracks of the carbon-carbon composite material and reacts with the carbon matrix or pyrolytic carbon to generate one or more ceramic phases of SiC, ZrC, HfC and VC.

[0067] Step 5: The Si-Zr-Hf-V reactive melt infiltration modified carbon-carbon composite material is subjected to hot isostatic pressing nitriding treatment to obtain the Si-Zr-Hf-VN modified carbon-carbon composite material.

[0068] The hot isostatic nitriding treatment is carried out in a nitrogen, ammonia, nitrogen-argon mixture, nitrogen-hydrogen mixture or nitrogen-containing atmosphere, with a nitriding temperature of 1600-2000℃, a nitriding pressure of 150-200MPa, and a holding time of 0.5-8h.

[0069] Example 2

[0070] This embodiment provides a method for preparing Si-Zr-Hf-VN modified carbon / carbon composite materials based on the screening results of Example 1 and verifying the mechanism.

[0071] First, according to the preferred Si-Zr-Hf-V quaternary reactive infiltration composition obtained in Example 1, Si, Zr, Hf, and V raw materials are weighed. The raw materials can be elemental powders, elemental particles, metal blocks, pre-alloyed powders, pre-alloyed blocks, or hydride powders. The raw materials are mixed according to a preferred atomic ratio to obtain the Si-Zr-Hf-V infiltration agent. The mixing method can be one or more of mechanical mixing, ball milling, tableting, cold pressing, melt alloying, or pre-alloying.

[0072] Then, the Si-Zr-Hf-V infiltrator was placed in contact with the carbon / carbon composite material and heated to 1450–2300 °C under a vacuum or argon protective atmosphere, and held at that temperature for 0.5–6 h. During the infiltration process, the Si-Zr-Hf-V infiltrator formed a molten or semi-molten state at high temperature and entered the open pores, fiber bundle gaps, and matrix microcracks of the carbon / carbon composite material under the drive of capillary action, wetting action, and interfacial chemical reactions. Si, Zr, Hf, and V reacted with the carbon matrix or pyrolytic carbon to generate SiC, ZrC, HfC, and VC ceramic phases, resulting in the Si-Zr-Hf-V reactive infiltrated modified carbon / carbon composite material.

[0073] Following reactive melting and infiltration, the Si-Zr-Hf-V reactively infiltrated modified carbon / carbon composite material is subjected to hot isostatic pressing (HIP) nitriding. The HIP nitriding is carried out in a nitrogen, ammonia, nitrogen-argon mixture, nitrogen-hydrogen mixture, or nitrogen-containing atmosphere, at a nitriding temperature of 1500–2000℃, a nitriding pressure of 150–200 MPa, and a holding time of 4–8 h.

[0074] During hot isostatic pressing (HIP) nitriding, nitrogen-containing gas enters the surface pores, residual open channels, and ceramic phase interface regions of the material under high temperature and isotropic pressure, promoting the diffusion of nitrogen atoms into the ZrC, HfC, and VC ceramic phases. Through the substitution of some carbon sites in the carbide lattice by nitrogen atoms, interstitial occupancy, or interfacial reactions, ZrC, HfC, and / or VC are at least partially transformed into... One or more of them, among which Thus, Si-Zr-Hf-VN modified carbon / carbon composite materials were obtained.

[0075] The Si-Zr-Hf-VN modified carbon / carbon composite material forms a multiphase ceramic protective structure containing SiC, ZrC, HfC, VC, carbonitrides, and / or nitrides inside and on the surface. Under high-temperature oxidative ablation conditions, SiC oxidizes to form… Or a Si-containing composite oxide viscous phase, used to spread and seal pores and cracks; ZrC, HfC, ZrN, HfN and their carbonitrides are oxidized to form , Alternatively, Zr / Hf-containing composite oxides can be used to form a high-melting-point solid-phase oxide framework; V, VC, VN, or V-based carbonitrides can be oxidized to generate V-based oxides or V-containing composite oxides to adjust the viscosity, flowability, or sintering behavior of the oxide layer. This results in an ablation-resistant protective structure with the synergistic effects of liquid-phase sealing, solid-phase framework support, and nitriding oxygen barrier.

[0076] Furthermore, to explain the mechanism by which hot isostatic pressing (HIP) nitriding improves the resistance to oxidation and ablation, this embodiment uses first-principles calculations to determine the diffusion barriers of oxygen atoms in carbides, carbonitrides, and nitrides. Specifically, the diffusion barriers of ZrC, HfC, VC, and... Crystal structure models of ZrN, HfN, and VN were developed, and their geometry was optimized. Oxygen atoms were introduced into the optimized crystal structures, and their initial and final stable positions were determined. The climbing elastic band method was used to calculate the diffusion path and diffusion barrier of oxygen atoms migrating from their initial to final stable positions. ; in, The oxygen atom diffusion barrier, This represents the total system energy corresponding to the highest energy transition state in the diffusion path. This represents the total energy of the system when the oxygen atom is in its initial stable position.

[0077] The diffusion barriers of oxygen atoms in ZrC, HfC, and VC are respectively compared with their... , , The diffusion barriers in ZrN, HfN, and VN were compared. If the oxygen diffusion barrier in the carbonitrides or nitrides formed after nitriding is higher than that in the corresponding carbides, it indicates that hot isostatic pressing nitriding can increase the migration resistance of oxygen atoms in the ceramic protective phase, reduce the diffusion rate of oxygen atoms into the carbon-carbon composite matrix, and thus improve the oxidation and ablation resistance of Si-Zr-Hf-VN modified carbon-carbon composites.

[0078] The above embodiments illustrate that the present invention can first establish a prediction model for density, linear ablation rate and mass ablation rate after melting and infiltration using experimental data of the Si-Zr-Hf-V whitelist system, then perform Pareto screening on the unknown Si-Zr-Hf-V quaternary composition space, and use the selected preferred composition for reactive melting and hot isostatic pressing nitriding preparation, and finally explain the oxygen barrier mechanism of nitriding treatment through first-principles calculations, thus realizing the integration of material composition design, process preparation and performance mechanism verification.

[0079] The method for preparing Si-Zr-Hf-V modified carbon-carbon composite materials provided in this embodiment utilizes experimental data from existing whitelist melt-infiltration systems to construct physically meaningful reactive melt-infiltration densification characteristics and ablation protection behavior characteristics. Depth prediction models, linear ablation rate prediction models, and mass ablation rate prediction models are established after melt-infiltration. Furthermore, batch prediction and Pareto optimization screening are performed in the unknown Si-Zr-Hf-V quaternary composition space to obtain a preferred composition with low linear and mass ablation rates. Simultaneously, reactive melt-infiltration preparation is carried out according to the preferred composition, and carbonitride or nitride protective phases are formed through hot isostatic pressing nitriding. The mechanism of nitriding-enhanced oxidation ablation resistance is explained by first-principles calculation of the oxygen diffusion barrier. This method enables integrated design from data collection, physical-inspired modeling, composition screening, material preparation to mechanism verification, which is of great significance for improving the service performance of carbon / carbon composite materials in extreme thermo-oxidative environments.

[0080] The above embodiments are preferred embodiments of this application. Those skilled in the art can make various changes or improvements based on them. Without departing from the overall concept of this application, these changes or improvements should fall within the scope of protection claimed in this application.

Claims

1. A method for preparing carbon-carbon composite materials with machine learning-assisted melt infiltration modification, characterized in that, include: Step 1: Construct several reactive melting candidate compositions. Using machine learning methods, based on the melting element composition information, pore structure parameters, preparation process parameters, preform density parameters, and ablation environment parameters of each candidate composition, predict the linear ablation rate and mass ablation rate of each candidate composition. Step 2: Using the predicted linear ablation rate and the predicted mass ablation rate as bi-objective optimization variables, extract the non-dominated solution set in the constructed bi-objective optimization space and extract the Pareto front. Step 3: Calculate the distance from each reactive infiltration candidate composition to the preset ideal point on the Pareto front based on the linear ablation rate and the mass ablation rate, and screen the optimal reactive infiltration composition based on the distance. Step 4: Prepare a melting agent according to the melting element composition information corresponding to the optimal reaction melting composition, and perform reaction melting treatment on the carbon-carbon composite matrix according to the preparation process parameters and ablation environment parameters corresponding to the optimal reaction melting composition.

2. The method for preparing carbon-carbon composite materials with machine learning-assisted melt infiltration modification according to claim 1, characterized in that, The candidate compositions for reactive infiltration were all selected from the Si-Zr-Hf-V infiltration system.

3. The method for preparing carbon-carbon composite materials with machine learning-assisted melt infiltration modification according to claim 2, characterized in that, After obtaining the Si-Zr-Hf-V reactive infiltration modified carbon-carbon composite material by performing reactive infiltration treatment in step 4, it is further subjected to hot isostatic pressing nitriding treatment to obtain the Si-Zr-Hf-VN modified carbon-carbon composite material.

4. The method for preparing carbon-carbon composite materials with machine learning-assisted melt infiltration modification according to claim 1, characterized in that, When extracting the non-dominated solution set in step 2, if the predicted linear ablation rate and the predicted mass ablation rate of a certain reaction infiltration candidate composition are both no greater than the other candidate composition in terms of both linear ablation rate and mass ablation rate, and at least one of the targets is less than the other candidate composition, then the former is considered to dominate the latter; all candidate compositions that are not dominated by other candidate compositions constitute the Pareto front.

5. The method for preparing carbon-carbon composite materials with machine learning-assisted melt infiltration modification according to claim 1, characterized in that, The preset ideal point in step 3 is The distance calculated based on the linear ablation rate and the mass ablation rate is expressed as follows: ； in, The distance from the candidate composition for reactive melt infiltration to the preset ideal point. and The linear ablation rate and mass ablation rate of the reactive infiltration candidate composition are respectively.

6. The method for preparing carbon-carbon composite materials with machine learning-assisted melt infiltration modification according to claim 1, characterized in that, The reactive infiltration treatment includes: The prepared infiltration agent was mixed and then placed in contact with the carbon-carbon composite matrix. The temperature is raised to 1400–2000℃ under vacuum, argon, helium, or a protective atmosphere and held for 0.5–4 hours to allow the infiltrator to enter the open pores, fiber bundle gaps, and matrix microcracks of the carbon-carbon composite matrix and react with the carbon matrix or pyrolytic carbon to generate one or more ceramic phases of SiC, ZrC, HfC, and VC.

7. The method for preparing carbon-carbon composite materials with machine learning-assisted melt infiltration modification according to claim 3, characterized in that, The hot isostatic nitriding treatment is carried out in a nitrogen, ammonia, nitrogen-argon mixture, nitrogen-hydrogen mixture or nitrogen-containing atmosphere, with a nitriding temperature of 1600-2000℃, a nitriding pressure of 150-200MPa, and a holding time of 0.5-8h.

8. The method for preparing carbon-carbon composite materials with machine learning-assisted melt infiltration modification according to claim 1, characterized in that, Machine learning methods are used to predict the linear ablation rate and mass ablation rate for each candidate component. Specifically: Based on the elemental composition information of the melt infiltration, pore structure parameters, preparation process parameters, and preform density parameters, a physical-inspired feature for characterizing the density after melt infiltration is constructed. Based on the predicted density after melting and infiltration, as well as information on the elemental composition of the melt infiltration, pore structure parameters, preparation process parameters, preform density parameters, and ablation environment parameters, physical heuristic features for characterizing ablation protection behavior are constructed. A density prediction model after melting and infiltration is constructed and trained using a training set. The input of the model is the physical heuristic features used to characterize the density after melting and infiltration, and the output is the density after melting and infiltration. A linear ablation rate prediction model and a mass ablation rate prediction model were constructed and trained using a training set. The inputs to the linear ablation rate prediction model and the mass ablation rate prediction model are both physical heuristic features used to characterize ablation protection behavior, and the outputs are linear ablation rate and mass ablation rate, respectively. Using a trained post-melting density prediction model, the corresponding post-melting density is predicted based on the physical heuristic features of each reactive melting candidate composition used to characterize ablation protection behavior. Then, using the trained linear ablation rate prediction model and mass ablation rate prediction model, the corresponding linear ablation rate and mass ablation rate are predicted respectively based on the physical heuristic features of each reactive infiltration candidate composition used to characterize the ablation protection behavior.

9. The method for preparing carbon-carbon composite materials with machine learning-assisted melt infiltration modification according to claim 8, characterized in that, The physical heuristic features used to characterize ablation protection behavior include one or more of the following: pyrolytic carbon deposition densification features, carbonization reaction enthalpy weighted features, transport-reaction coupling features, and penetration driving force features; The pyrolysis carbon deposition densification characteristics are characterized by calculating the following features: pyrolysis carbon deposition increment characteristics. Relative densification characteristics of pyrolytic carbon Pyrolysis carbon filling efficiency characteristics : ； ； ； In the formula, Density before deposition Density after deposition Open area ratio; The enthalpy-weighted characteristic of the carbonization reaction is characterized by calculating the following feature: the total component carbonization reaction enthalpy weighted value. Normalized total component carbonization reaction enthalpy characteristics Enthalpy-weighted values ​​of carbonization reactions of non-Cu active carbide elements Average enthalpy characteristics of non-Cu active carbide elements Ratio of active carbide forming elements : ； ； ； ； ； In the formula, For infiltration elements atomic ratio fraction, For infiltration elements Enthalpy of carbonization reaction; The element is Cu; The transport-reaction coupling characteristic is a composite characteristic calculated from porosity, carbonization volume expansion, average atomic radius of active elements, melting point of the corresponding carbide of active elements, and preparation temperature. ； In the formula, It is a composite feature; The weighted average of the carbonization volume expansion of multi-element modification for high-temperature ceramics. This is the weighted average of the average atomic radii of non-Cu reactive elements. This is the weighted average of the melting points of the carbides corresponding to non-Cu reactive elements. For preparation temperature; The permeation driving force characteristic is denoted as The average cohesive energy, average electronegativity difference, average atomic weight, and average melting point were calculated based on the weighted average atomic ratios of the infiltrating elements. ； In the formula, The average cohesive energy is obtained by weighting the atomic ratio of the infiltrating elements. The average electronegativity difference of elements relative to carbon atoms, obtained by weighting according to the atomic ratio. The average atomic weight is obtained by weighting the atomic ratios of the elements. The average melting point is obtained by weighting the atomic ratios of the elements; where the average cohesive energy is... Poor average electronegativity Average atomic weight Average melting point They are uniformly denoted as equivalent physical properties. The weighted average of the physical properties of each element is obtained based on the atomic ratio of the infiltrating elements. ； In the formula, For infiltration elements atomic ratio, For infiltration elements Property characteristics to be determined The corresponding physical property parameters.

10. The method for preparing carbon-carbon composite materials with machine learning-assisted melt infiltration modification according to claim 8, characterized in that, The physical heuristic features used to characterize ablation protection behavior include one or more of the following: basic protection features, liquid phase protection features, solid phase protection features, and liquid-solid equilibrium features. The basic protective features are used to characterize the coupling effect of the material's initial skeleton density, the degree of melt infiltration densification retention, and the effectiveness of pore filling. The calculation formula is as follows: ； ； ； ； in, Represents basic protection features. Indicates density before deposition. For melt penetration retention rate, To improve melt infiltration filling efficiency, This is the predicted density value after melt infiltration. Density after CVI The equivalent mixing density of the melt-infiltrated phase, For infiltration elements mass fraction, For element infiltration metal density, Indicates the open area ratio; The liquid phase protection feature is used to characterize the formation, flow, spreading, and plugging ability of liquid or viscous oxidation products under high-temperature ablation conditions. The calculation formula is as follows: ； ； ； ； ； in, Represents liquid phase protection characteristics, This represents the total volume of the liquid-phase oxidation products. The average optical alkalinity in the liquid phase. The average ion field strength in the liquid phase; For the normalized characteristics of heat load, For the original heat load, This is the average of the heat load samples. Oxygen flow rate, For gas flow rate, For ablation time, This refers to the distance between the spray gun and the spray gun. This represents the normalized characteristic of the oxidative load. For the original oxidation load, This represents the mean of the oxidative load samples. The solid-phase protection features are used to characterize the volume reserve, thermal stability, and resistance to high-temperature instability of the high-melting-point solid-phase oxide framework, and the calculation formula is as follows: ； in, Represents solid-phase protection characteristics. This represents the total volume of the solid-phase oxidation products. This is an indicator of solid-phase stability. The liquid-solid equilibrium characteristic is used to characterize the relative dominance of the liquid-phase protection mechanism and the solid-phase protection mechanism, and the calculation formula is: ； in, Represents liquid-solid equilibrium characteristics. To prevent extremely small positive numbers with a denominator of zero.

Images