Preparation method of aerogel two-component foaming fireproof plugging material

Abstract

The invention provides a preparation method of an aerogel two-component foaming fireproof plugging material, and belongs to the technical field of computational chemistry, and the preparation method comprises the following steps: firstly preparing aerogel sol and loading a carbon monoxide catalyst and a flame retardant, and then preparing a component A (polyhydric alcohol, the catalyst, the flame retardant and a foaming agent) containing aerogel and a component B (polyisocyanate and aerogel) containing aerogel. Optimal process parameters are predicted through an aerogel structure optimization network model, and the two components are mixed according to the mass ratio of 1: 1-1: 1.5 for reaction and foaming forming. The foaming process follows a flame propagation inhibition kinetics, heat conduction barrier and aerogel porosity and heat insulation performance relation equation. And after the performance of the finished product is tested, the attention weight distribution of the network model is adjusted by using deviation data, so that online optimization and updating are realized, and the problem of collaboration of heat insulation and toxic gas inhibition of the fireproof plugging material is solved.

Description

Technical Field

[0001] This invention belongs to the field of computational chemistry technology, and specifically relates to a method for preparing a two-component aerogel foam fireproof sealing material. Background Technology

[0002] Fire-stopping materials are widely used for fireproofing cable and pipe penetrations through walls and for fire isolation of various equipment in buildings, and are an important component of modern building fire safety systems. Traditional fire-stopping materials mainly include inorganic fire-stopping materials, organic fire-stopping materials, and composite fire-stopping materials. These materials achieve fireproofing effects through physical isolation, flame-retardant additives, and intumescent components. In commercial buildings, industrial facilities, and high-rise residential buildings, fire-stopping materials are often used to fill holes and gaps in walls and floors to create fire-resistant zones and prevent the spread of fire.

[0003] However, traditional fire-stopping materials generally suffer from drawbacks such as insufficient thermal insulation, high density, and low strength. In particular, they are prone to releasing toxic gases such as carbon monoxide under high-temperature conditions, posing a serious threat to personnel evacuation and fire rescue. While organic fire-stopping materials possess good elasticity and adhesion, their fire resistance limit is relatively low; inorganic fire-stopping materials have higher fire resistance, but poor seismic and crack resistance; composite fire-stopping materials, although improving some properties to a certain extent, lack structural stability and an effective mechanism for handling toxic gases during high-temperature decomposition.

[0004] Currently, the core challenge facing fire-stopping material technology is how to effectively suppress toxic gases such as carbon monoxide produced during a fire while ensuring low thermal conductivity and high fire resistance, and also ensuring that the material has sufficient mechanical strength and structural stability. In other words, existing technologies struggle to simultaneously achieve highly efficient thermal insulation and toxic gas suppression in fire-stopping materials. Summary of the Invention

[0005] In view of this, the present invention provides a method for preparing a two-component aerogel foamed fireproof sealing material, which can solve the technical problem of fireproof sealing materials in the prior art that are difficult to achieve both high-efficiency heat insulation and toxic gas suppression at the same time.

[0006] This invention is achieved as follows: A method for preparing a two-component aerogel foamed fire-resistant sealing material includes: preparing an aerogel sol; loading a carbon monoxide catalyst and a flame retardant into the aerogel sol; formulating component A containing the aerogel-loaded carbon monoxide catalyst and the aerogel-loaded flame retardant, and component B containing aerogel; using an aerogel structure optimization network model to predict and determine optimal process parameters; mixing and reacting component A and component B according to the predicted process parameters to form a foam, wherein the foaming process follows the flame propagation inhibition kinetic equation, the heat conduction barrier equation, and the equation relating aerogel porosity to thermal insulation performance; conducting performance tests on the product and obtaining performance deviation data; using the performance deviation data to optimize the aerogel structure optimization network model and redetermine the process parameters.

[0007] Specifically, the step of preparing aerogel sol involves selecting one of silica aerogel, carbon aerogel, or alumina aerogel as a raw material and using the sol-gel method to prepare an aerogel sol with a particle size of 1 to 100 micrometers. The aerogel has a three-dimensional network structure and low density and high porosity.

[0008] Specifically, the step of supporting the carbon monoxide catalyst involves adding a hogalat agent to an aerogel sol and dispersing it evenly, adding a crosslinking agent to make the sol form a gel state, and then aging, surface modification, and gradient vacuum drying to obtain the aerogel-supported carbon monoxide catalyst.

[0009] Specifically, the step of loading the flame retardant involves adding one or more flame retardants selected from ammonium polyphosphate, expanded graphite, glass microspheres, or montmorillonite to an aerogel sol and dispersing them evenly. A crosslinking agent is then added to form a gel state in the sol. After aging, surface modification, and gradient vacuum drying, the aerogel-loaded flame retardant is obtained.

[0010] Specifically, the step of preparing component A involves mixing 60 to 85% by mass of a polyol, 0.1 to 3% by mass of an aerogel-supported carbon monoxide catalyst, 10 to 20% by mass of an aerogel-supported flame retardant, and 2 to 10% by mass of a foaming agent to obtain component A. The foaming agent is one of dichlorofluoroethane or cyclopentane.

[0011] Specifically, the step of preparing component B involves mixing 90 to 95% by mass of polyisocyanate with 5 to 10% by mass of aerogel in a uniform mass ratio to obtain component B.

[0012] The mass ratio of component A to component B is 1:1 to 1:1.5, which ensures that the material achieves optimal crosslinking degree and physical properties during the curing reaction.

[0013] The performance deviation data refers to the difference between the measured actual performance indicators and the expected target performance indicators, including density deviation, compressive strength deviation, thermal conductivity deviation, and fire resistance limit deviation.

[0014] The specific structure of the aerogel structure optimization network model is a deep learning model combining multilayer convolutional neural networks and graph neural networks, including an input layer, a feature extraction layer, a structure prediction layer, and an output layer. The feature extraction layer uses a three-dimensional convolutional neural network to extract the spatial features of the aerogel material, the structure prediction layer uses a graph attention network to capture the key connection points in the aerogel skeleton structure, and the output layer predicts the three-dimensional network structure characteristics of the aerogel and its impact on material properties.

[0015] Specifically, the optimization steps of the aerogel structure optimization network model involve calculating performance deviation weights using performance deviation data, inputting these weights into the attention mechanism of the aerogel structure optimization network model, adjusting the distribution of attention weights in the aerogel structure optimization network model, and performing online optimization and updates to the aerogel structure optimization network model.

[0016] The flame propagation suppression kinetic equation described above is used to describe the flame propagation suppression process in flame-retardant materials. The inputs include material density, aerogel content, aerogel-loaded flame retardant concentration, temperature gradient, and oxygen concentration. The outputs are flame propagation rate and heat release rate.

[0017] The thermal conductivity barrier equation is used to calculate the thermal conductivity of aerogel composites. The inputs include aerogel content, porosity, aerogel particle size distribution, material density, and temperature gradient. The output is the effective thermal conductivity of the material. The equation relating aerogel porosity to thermal insulation performance is used to establish the mathematical relationship between the pore structure of aerogel materials and their thermal insulation effect. The inputs include the average pore size of the aerogel, porosity, thermal conductivity of the framework material, gas thermal conductivity, and radiative thermal conductivity. The output is the comprehensive thermal conductivity coefficient.

[0018] The aging process refers to the process of allowing the gel to stand under specific temperature and time conditions to further stabilize its internal structure; the surface modification process refers to the process of changing the surface properties of the aerogel through chemical methods to improve its hydrophobic properties and stability; the gradient decompression drying process reduces the capillary force during the evaporation of liquid in the gel by gradually reducing the pressure, preventing the aerogel structure from collapsing and maintaining its porous characteristics.

[0019] The process parameters mentioned include mixing temperature, stirring rate, molding pressure, and curing time.

[0020] The performance tests include testing the prepared aerogel two-component foamed fireproof sealing material for performance indicators such as density, compressive strength, thermal conductivity and fire resistance limit, to ensure that the density is 120 to 180 kg per cubic meter, the compressive strength is not less than 1.0 MPa, the thermal conductivity is not higher than 0.040 W / m Kelvin, and the fire resistance limit is not less than 2 hours.

[0021] This invention combines aerogel with a polyurethane foam system, and introduces horgalactone and flame retardants to construct a multifunctional synergistic fire-stopping material system. This material fully utilizes the three-dimensional network structure and nanoscale porosity of the aerogel, significantly reducing the material's thermal conductivity to no more than 0.040 W / m Kelvin, increasing the fire resistance limit to no less than 2 hours, while maintaining a low density (120 to 180 kg / m³) and sufficient compressive strength (no less than 1.0 MPa).

[0022] This invention establishes a mechanism for the conversion of toxic gases during a fire through an innovative design of aerogel-supported carbon monoxide catalysts. This effectively solves the problem that traditional fireproof materials only focus on physical insulation while neglecting the treatment of toxic gases. The high porosity and large specific surface area of ​​the aerogel provide an ideal loading platform for the catalyst, enabling the material to actively catalyze the decomposition of carbon monoxide into carbon dioxide under high-temperature conditions, significantly reducing the risk of poisoning in a fire.

[0023] This invention applies a deep learning aerogel structure optimization network model to predict and optimize process parameters online, realizing a precise mapping relationship between the microstructure of aerogel and its macroscopic fireproof performance. It solves the problem of relying on empirical trial and error and lacking theoretical guidance in the development of traditional fireproof sealing materials, ensuring that the product performance is stable and controllable and can adapt to the needs of different application scenarios. It successfully solves the technical problem of the synergistic effect of high-efficiency heat insulation and toxic gas suppression. Attached Figure Description

[0024] Figure 1 This is a flowchart of the method of the present invention.

[0025] Figure 2 This is a comparison diagram of the microstructure of the aerogel generated before and after the optimization of process parameters in Example 2. Detailed Implementation

[0026] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings.

[0027] like Figure 1 The diagram shown is a flowchart of a method for preparing a two-component aerogel foamed fireproof sealing material provided by the present invention. This method includes the following steps:

[0028] S01. Preparation of aerogel sol: Select one of silica aerogel, carbon aerogel or alumina aerogel as raw material, and prepare aerogel sol with a particle size of 1 to 100 micrometers by sol-gel method. The aerogel has a three-dimensional network structure and low density and high porosity.

[0029] S02, Supported carbon monoxide catalyst: Hogalat agent is added to the aerogel sol prepared in step S01 and dispersed evenly. A crosslinking agent is added to make the sol form a gel state. After aging, surface modification and gradient vacuum drying, the aerogel supported carbon monoxide catalyst is obtained.

[0030] S03, Flame retardant loaded: One or more flame retardants selected from ammonium polyphosphate, expanded graphite, glass microspheres or montmorillonite are added to the aerogel sol prepared in step S01 and dispersed evenly. A crosslinking agent is added to make the sol form a gel state. After aging, surface modification and gradient vacuum drying, the aerogel loaded flame retardant is obtained.

[0031] S04. Prepare component A by mixing 60 to 85% by mass of polyol, 0.1 to 3% by mass of aerogel-supported carbon monoxide catalyst, 10 to 20% by mass of aerogel-supported flame retardant, and 2 to 10% by mass of foaming agent evenly to obtain component A. The foaming agent is one of dichlorofluoroethane or cyclopentane.

[0032] S05. Prepare component B by mixing 90 to 95% by mass of polyisocyanate and 5 to 10% by mass of aerogel in a uniform mass ratio to obtain component B.

[0033] S06. Determination of process parameters: The three-dimensional network structure of aerogel is predicted and analyzed using the aerogel structure optimization network model. The optimal process parameters are determined based on the prediction results, including mixing temperature, stirring rate, molding pressure and curing time.

[0034] S07. Mixing reaction: Mix component A and component B according to the process parameters determined in step S06, stir thoroughly to allow them to react, pour into a preset mold, and solidify and foam at room temperature. The foaming process follows the flame propagation inhibition kinetic equation, the heat conduction barrier equation, and the equation relating aerogel porosity to thermal insulation performance.

[0035] S08. Performance testing: The prepared aerogel two-component foamed fireproof sealing material is tested for performance indicators such as density, compressive strength, thermal conductivity, and fire resistance limit to ensure that the density is 120 to 180 kg / m³, the compressive strength is not less than 1.0 MPa, the thermal conductivity is not higher than 0.040 W / m Kelvin, and the fire resistance limit is not less than 2 hours. The performance deviation data between the test results of the aforementioned performance indicators and the target performance are calculated.

[0036] S09. Network model optimization: Calculate the performance deviation weights using the performance deviation data obtained in step S08, input the performance deviation weights into the attention mechanism of the aerogel structure optimization network model, adjust the attention weight distribution in the aerogel structure optimization network model, and perform online optimization and update of the aerogel structure optimization network model.

[0037] S10. Quality control: Based on the optimized aerogel structure network model optimized in step S09, re-predict the optimal process parameters, adjust the preparation process, and ensure that the product performance is stable and meets the fireproof sealing requirements.

[0038] Among them, hogalat is a catalyst used to catalyze the conversion of carbon monoxide into carbon dioxide, which can effectively eliminate toxic carbon monoxide gas produced in fires.

[0039] Aging refers to the process of allowing a gel to stand under specific temperature and time conditions to further stabilize its internal structure.

[0040] Among them, surface modification refers to the process of changing the surface properties of aerogels through chemical methods to improve their hydrophobicity and stability.

[0041] Gradient pressure reduction drying is a special drying technique that gradually reduces the pressure to decrease the capillary force during the evaporation of liquid in the gel, preventing the aerogel structure from collapsing and maintaining its porous properties.

[0042] The three-dimensional network structure refers to the interconnected three-dimensional skeleton structure inside the aerogel, forming a large number of nanoscale pores. The three-dimensional network structure is the basis for the aerogel's low density and low thermal conductivity characteristics.

[0043] The ratio of component A to component B is based on the specific ratio described in Chinese invention patent application number 202510566817.3. Specifically, the mass ratio of component A to component B is determined by calculating the functional group equivalent ratio of polyol to polyisocyanate in component A (60 to 85% by mass), aerogel-supported carbon monoxide catalyst (0.1 to 3% by mass), foaming agent (2 to 10% by mass), and aerogel-supported flame retardant (10 to 20% by mass) and polyisocyanate (90 to 95% by mass) and aerogel in component B (5 to 10% by mass). The final mass ratio of component A to component B is determined to be 1:1 to 1:1.5. This mass ratio ensures that the material achieves optimal crosslinking degree and physical properties during the curing reaction.

[0044] The flame propagation suppression kinetic equation describes the flame propagation suppression process in flame-retardant materials and provides guidance for predicting the material's behavior under fire conditions. The inputs include material density, aerogel content, aerogel-loaded flame retardant concentration, temperature gradient, and oxygen concentration from steps S04 and S05. The outputs are the flame propagation rate and heat release rate, which are used for optimizing process parameters in step S06. The flame propagation suppression kinetic equation expresses the relationship between material composition and flame propagation rate, providing a theoretical basis for optimizing material formulations.

[0045] The thermal conductivity barrier equation is used to calculate the thermal conductivity of aerogel composites, describing the barrier effect of heat transfer in the material. The inputs include the aerogel content, porosity, aerogel particle size distribution, material density, and temperature gradient obtained in steps S01 to S05. The output is the effective thermal conductivity of the material, which is used for performance evaluation in step S08. The thermal conductivity barrier equation reveals the relationship between the microstructure of aerogel and its macroscopic thermal insulation performance, providing theoretical guidance for improving the thermal insulation effect of materials.

[0046] The equation relating aerogel porosity to thermal insulation performance is used to establish the mathematical relationship between the pore structure of aerogel materials and their thermal insulation effect. The inputs include the average pore size, porosity, thermal conductivity of the framework material, gas thermal conductivity, and radiative thermal conductivity of the aerogel prepared in step S01. The output is the comprehensive thermal conductivity coefficient, which is used to optimize the process parameters in step S06. The equation relating aerogel porosity to thermal insulation performance explains the physical mechanism of the excellent thermal insulation performance of aerogel and provides a basis for the structural design of aerogel materials.

[0047] Among them, the performance deviation data refers to the difference between the actual performance index measured in step S08 and the expected target performance index, including density deviation, compressive strength deviation, thermal conductivity deviation and fire resistance limit deviation.

[0048] Among them, the performance deviation weight is a value obtained by weighting the performance deviation data according to the importance of different performance indicators, and is used to quantify the degree of influence of different performance indicators on the final product performance.

[0049] Among them, the attention mechanism is a key module in the aerogel structure optimization network model used to learn the importance of input features, and optimizes the model prediction accuracy by adjusting the weight distribution of different features.

[0050] The specific structure of the aerogel structure optimization network model is a deep learning model combining multi-layer convolutional neural networks and graph neural networks, including an input layer, a feature extraction layer, a structure prediction layer, and an output layer. The feature extraction layer uses a three-dimensional convolutional neural network to extract the spatial features of the aerogel material, the structure prediction layer uses a graph attention network to capture key connection points in the aerogel skeleton structure, and the output layer predicts the three-dimensional network structure characteristics of the aerogel and its impact on material properties. The steps for establishing the training dataset for the aerogel structure optimization network model specifically include collecting three-dimensional network structure data of aerogel samples under different preparation conditions, standardizing the three-dimensional network structure data, extracting key feature parameters, and establishing a mapping relationship between preparation process parameters and three-dimensional network structure characteristics. The system employs data augmentation techniques to expand the training samples, forming a comprehensive dataset containing raw material composition, process condition parameters, and structural characteristic parameters. The training steps for the aerogel structure optimization network model specifically include initializing network parameters, optimizing the aerogel structure optimization network model using a mini-batch stochastic gradient descent algorithm, determining the optimal hyperparameters through cross-validation, using an early stopping strategy to prevent overfitting, and fine-tuning the aerogel structure optimization network model based on material performance test results. Ultimately, this forms a deep learning model capable of inversely deriving the optimal process parameters based on desired material properties. The online optimization update refers to the process of continuously adjusting the network model parameters based on actual test data during production, without retraining the entire model, thus improving the model's adaptability to new data.

[0051] The specific implementation methods of the above steps are described in detail below.

[0052] The specific implementation of step S01 involves selecting one of silica aerogel, carbon aerogel, or alumina aerogel as a raw material and preparing an aerogel sol with a particle size of 1 to 100 micrometers using the sol-gel method. Specifically, a silicon precursor, such as tetraethyl orthosilicate or tetramethoxysilane, is first dissolved in ethanol. Deionized water is added for hydrolysis, and a catalyst, such as ammonia, is added to adjust the pH to 8-9. The reaction temperature is controlled at 25-30°C, and the reaction time is 4-6 hours to ensure complete hydrolysis of the silicon precursor. Then, a condensation reaction is carried out at 50-60°C for 6-8 hours to form a sol system with a preliminary network structure. The sol particle size is monitored in real time using dynamic light scattering technology, and the reaction is stopped when it reaches the 1-100 micrometer range. The resulting sol is then treated with ultrasonic dispersion technology for 10-15 minutes to ensure uniform particle size distribution. The purpose of this step is to prepare an aerogel sol with a three-dimensional network structure, low density, and high porosity. This structure is the basis for subsequent loading of functional materials and achieving excellent fire-retardant properties.

[0053] The specific implementation of step S02 involves adding the hogalat agent to the aerogel sol prepared in step S01 to prepare an aerogel-supported carbon monoxide catalyst via a loading process. Specifically, the hogalat agent is first prepared into an aqueous solution at a ratio of 3-5% by mass of the aerogel, and the pH value is controlled within the range of 5.5-6.5 using a pH adjustment method. Then, the hogalat agent aqueous solution is mixed with the aerogel sol using an ultrasonic-assisted impregnation method. The ultrasonic power is set to 300-400 watts for 20-30 minutes to ensure that the hogalat agent is uniformly dispersed in the aerogel sol. Next, a crosslinking agent such as dimethyldimethoxysilane is added at a concentration controlled at 0.5-1.5%. After stirring evenly, the mixture is allowed to stand for 12-24 hours to allow the sol to transform into a gel state. The gel is then aged at 60-70°C for 48-72 hours to further stabilize its internal structure. Subsequently, surface modification was performed using trimethylchlorosilane as a modifier. The aerogel was treated with vapor phase deposition at 80–90°C for 6–8 hours to improve its hydrophobic properties. Finally, a gradient vacuum drying process was conducted, starting at atmospheric pressure and gradually decreasing to 2–5 kPa at a rate of 10–15 kPa per hour, while maintaining the temperature at 50–60°C for 24–36 hours, to obtain the aerogel-supported carbon monoxide catalyst. The purpose of this step is to ensure that the hogalat agent is uniformly distributed within the three-dimensional network structure of the aerogel, achieving highly efficient catalytic oxidation of carbon monoxide generated in fires.

[0054] The specific implementation of step S03 involves adding a flame retardant to the aerogel sol prepared in step S01, and preparing an aerogel-loaded flame retardant through a loading process. Specifically, one or more flame retardants selected from ammonium polyphosphate, expanded graphite, glass microspheres, or montmorillonite are prepared as a suspension at a ratio of 15-20% of the aerogel mass. The flame retardant suspension is mixed with the aerogel sol using high-shear emulsification technology, with the shear rate controlled at 8000-10000 rpm for 15-20 minutes to ensure uniform dispersion of the flame retardant in the aerogel sol. A crosslinking agent, such as tetraethoxysilane, is added at a concentration controlled at 1-2%, and after stirring evenly, the mixture is allowed to stand for 18-24 hours to allow the sol to transform into a gel state. The gel is then aged at 65-75°C for 60-72 hours to fully stabilize its internal structure. Surface modification was performed using hexyltrimethoxysilane as a modifier at a concentration of 3–5%. The material was treated with a liquid-phase impregnation method at 75–85°C for 8–12 hours to improve its water resistance and stability. Finally, a gradient decompression drying process was conducted. The initial pressure was atmospheric pressure, gradually reduced to 3–6 kPa at a rate of 8–12 kPa per hour, while maintaining the temperature at 55–65°C for 30–40 hours, resulting in an aerogel-supported flame retardant. The purpose of this step was to ensure the flame retardant was uniformly distributed within the three-dimensional network structure of the aerogel, enabling it to exert multiple flame-retardant effects under fire conditions, including hindering heat conduction, absorbing heat, and forming a protective layer.

[0055] The specific implementation of step S04 involves preparing component A by uniformly mixing polyol, aerogel-supported carbon monoxide catalyst, aerogel-supported flame retardant, and foaming agent in a specific ratio. Specifically, a polyether polyol is selected as the main component, with a molecular weight range of 3000–6000, a hydroxyl value of 350–450 mg / g potassium hydroxide, and a content of 60–85% by mass. 0.1–3% by mass of the aerogel-supported carbon monoxide catalyst prepared in step S02 and 10–20% by mass of the aerogel-supported flame retardant prepared in step S03 are added. A foaming agent such as dichlorofluoroethane or cyclopentane is selected, with a content of 2–10% by mass. Mixing is performed using a high-speed disperser at a speed of 1200–1500 rpm for 15–20 minutes, with the temperature controlled at 20–25°C. During the mixing process, particle size analysis technology is used to monitor the dispersion state in real time to ensure uniform distribution of each component and no obvious agglomeration. After mixing, vacuum degassing is performed at 50–80 kPa for 10–15 minutes to remove air bubbles introduced during mixing. The prepared component A is then sealed and stored below 20°C to prevent moisture and air contamination. The purpose of this step is to mix the polyol with hydroxyl functional groups and functional additives to form a homogeneous component A, providing a basis for subsequent reaction with component B.

[0056] The specific implementation of step S05 involves preparing component B by uniformly mixing polyisocyanate and aerogel in a specific ratio. Specifically, polymeric MDI (diphenylmethane diisocyanate) is selected as the polyisocyanate, with an NCO content of 30-32% and a mass percentage of 90-95%. The original aerogel prepared in step S01 is added, with a mass percentage of 5-10%. Low-shear mixing technology is used, with a rotation speed controlled at 600-800 rpm for 10-15 minutes and a temperature strictly controlled at 15-20°C to prevent premature reaction of the polyisocyanate. During mixing, infrared spectroscopy is used to monitor the change in isocyanate group content in real time to ensure the stable activity of component B. Immediately after mixing, vacuum degassing is performed, reducing the pressure to 60-90 kPa for 8-12 minutes to remove air bubbles and moisture introduced during mixing. The prepared component B must be sealed and stored in a dry environment at a temperature controlled at 10-15°C to avoid contact with moisture. The purpose of this step is to prepare component B containing isocyanate functional groups, wherein the added aerogel can enhance the mechanical and fire-resistant properties of the final product.

[0057] The specific implementation of step S06 involves using an aerogel structure optimization network model to predict and analyze the three-dimensional network structure of the aerogel and determine the optimal process parameters. Specifically, the raw material composition, proportioning data, and preparation conditions from steps S01 to S05 are first input into the aerogel structure optimization network model. This model, based on deep learning principles, employs an architecture combining a three-dimensional convolutional neural network and a graph attention network, enabling it to predict the microstructural characteristics of aerogels under different conditions. During model analysis, key features such as aerogel skeleton connection points, pore distribution, and density gradient are extracted and analyzed to calculate the optimal combination of process parameters. Based on the model output, the mixing temperature is determined to be 23–27°C, the stirring rate 1800–2200 rpm, the molding pressure 0.15–0.25 MPa, and the curing time 24–36 hours. The model also uses Monte Carlo simulation to predict product performance under different parameter combinations and selects the optimal parameter combination. The purpose of this step is to use a computer model to assist in determining the optimal process parameters, improving the stability and predictability of product performance, and reducing the number of experiments and material waste.

[0058] The specific implementation of step S07 involves mixing component A and component B according to the process parameters determined in step S06 to achieve foaming and molding of the material. Specifically, component A and component B are precisely weighed at a mass ratio of 1:1 to 1:1.5, and the optimal ratio is determined based on the functional group equivalent ratio, ensuring the isocyanate index is within the range of 1.05 to 1.15. Under the temperature conditions determined in step S06, a high-efficiency turbine mixer is used to mix the two components at a determined stirring rate, with the mixing time controlled at 10 to 15 seconds to ensure thorough homogeneity without over-reaction. The mixture is immediately poured into a pre-treated mold, which is preheated to 25 to 30°C, and a release agent is applied to the surface. The material is held at a determined molding pressure for a fixed time to allow it to fully foam and fill the mold cavity. During the foaming process, the material properties are monitored and predicted based on the flame propagation inhibition kinetics equation, the heat conduction barrier equation, and the equation relating aerogel porosity to thermal insulation performance. The curing time is set according to the model prediction results and is carried out at room temperature with humidity controlled at 40% to 60%. After curing, the material undergoes demolding and post-processing, including surface finishing and dimensional inspection. The purpose of this step is to control the polyurethane chemical reaction between components A and B, and to combine this with the properties of aerogel to form a fire-resistant sealing material with a closed-cell structure.

[0059] The specific implementation of step S08 involves conducting performance tests on the prepared aerogel two-component foamed fireproof sealing material to ensure it meets relevant technical specifications. Specifically, the material density is first tested according to GB / T6343 standard, using Archimedes' principle. The sample size is 100×100×100 mm, and the measurement temperature is 23±2℃. The required density is controlled between 120 and 180 kg / m³. The compressive strength is then tested according to GB / T8813 standard, using a sample size of 50×50×50 mm, a compression rate of 5 mm / min, and a test temperature of 23±2℃. The required compressive strength is not less than 1.0 MPa. Finally, the thermal conductivity is tested according to GB / T10295 standard, using the heat flow meter method. The sample size is 300×300×25 mm, and the test temperature is 25℃. The required thermal conductivity is not higher than 0.040 W / m Kelvin. The fire resistance limit is tested according to GB / T9978 standard. The sample size is determined based on the actual application scenario, typically 1000×1000×100 mm, and the fire resistance limit is required to be no less than 2 hours. After the test, the deviation values ​​between each performance index and the target performance are calculated, including density deviation not exceeding ±5%, compressive strength deviation not exceeding ±8%, thermal conductivity deviation not exceeding ±10%, and fire resistance limit deviation not exceeding ±15%. The purpose of this step is to comprehensively evaluate the physical and mechanical properties and fire resistance performance of the material, providing data support for subsequent optimization.

[0060] The specific implementation of step S09 involves calculating performance deviation weights using the performance deviation data obtained in step S08 to optimize the aerogel structure optimization network model. Specifically, the deviations of each performance index are first normalized, converting deviation data of different dimensions into a unified scale. A multi-criteria decision-making method, such as the analytic hierarchy process (AHP), is used to determine the relative importance of each performance index. For fire-resistant sealing materials, the weights are typically 0.4 for fire resistance limit, 0.3 for thermal conductivity, 0.2 for density, and 0.1 for compressive strength. Based on these weights and the normalized deviation values, the comprehensive performance deviation weights are calculated. The calculated performance deviation weights are input into the attention mechanism of the aerogel structure optimization network model, and the backpropagation algorithm is used to adjust the distribution of attention weights in the model. The network uses an adaptive learning rate, initially set to 0.001, dynamically adjusted according to changes in the loss function. An early stopping strategy is employed during optimization; training stops if the validation set loss shows no improvement after five consecutive iterations to prevent overfitting. The final optimized aerogel structure optimization network model is obtained, with its prediction accuracy improved by 15-20% compared to the original model. The purpose of this step is to optimize and update the theoretical model online using actual production data, so that the model's predictions are more consistent with actual production conditions.

[0061] The specific implementation of step S10 involves re-predicting the optimal process parameters based on the optimized aerogel structure network model from step S09, and then performing quality control. Specifically, the current production conditions and raw material parameters are first input, and the optimized aerogel structure network model is run to obtain new process parameter prediction results. Based on the prediction results, the preparation process is adjusted, including adjusting the mixing temperature range by ±2℃, the stirring rate range by ±200 rpm, the molding pressure range by ±0.05 MPa, and the curing time range by ±4 hours. Small-batch trial production is then conducted using the adjusted process parameters, with at least three samples taken from each batch, and the performance test from step S08 is repeated. If the test results meet the technical requirements, production scale is expanded; if deviations still exist, steps S09 to S10 are repeated until the product performance is stable and meets the requirements. Simultaneously, a production process quality control system is established, including raw material incoming inspection, process parameter monitoring, and product outgoing inspection, to ensure stable and reliable product quality. The purpose of this step is to continuously improve product performance and production processes through a closed-loop feedback optimization system, ensuring stable product performance and meeting fire-stopping requirements.

[0062] The detailed implementation of the aerogel structure optimization network model involves constructing a deep learning model combining multi-layer convolutional neural networks and graph neural networks. This model comprises four main parts: an input layer, a feature extraction layer, a structure prediction layer, and an output layer. The input layer receives parameters such as raw material composition and processing conditions, forming a 64×1 vector containing features such as aerogel type, particle size distribution, porosity, density, and surface modification degree. The feature extraction layer employs a three-dimensional convolutional neural network, consisting of three convolutional blocks. Each convolutional block contains a three-dimensional convolutional layer, a batch normalization layer, a ReLU activation function layer, and a max-pooling layer. The first convolutional layer uses 32 3×3×3 convolutional kernels with a stride of 1 and padding of 1; the second convolutional layer uses 64 3×3×3 convolutional kernels with a stride of 1 and padding of 1; and the third convolutional layer uses 128 3×3×3 convolutional kernels with a stride of 1 and padding of 1. The output of the feature extraction layer is a 128×8×8×8 feature map containing spatial feature information of the aerogel material. The structure prediction layer employs a graph attention network. First, the feature map is converted into graph structure data, where nodes represent intersections in the aerogel skeleton and edges represent skeleton connections. The graph attention network contains three graph attention layers, each with eight attention heads. The hidden layer has a dimension of 64 and uses LeakyReLU as the activation function. Cosine similarity is used to calculate the attention coefficients. The final layer normalizes the attention weights using the Softmax function. The output layer maps the output features of the graph attention network to the target parameter space through a fully connected network. It contains two fully connected layers: the first layer has 256 neurons, and the second layer has 128 neurons. A dropout rate of 0.3 is used to prevent overfitting. The final output layer has 16 neurons, corresponding to the predicted process parameters and structural characteristics.

[0063] The detailed steps for establishing the training dataset for the aerogel structure optimization network model involve the systematic collection and processing of aerogel structure and performance data. First, three-dimensional network structure data of aerogel samples under different preparation conditions are collected. Microscopic morphology images of the aerogels are obtained using techniques such as scanning electron microscopy, transmission electron microscopy, and X-ray tomography, with an image resolution of at least 10 nanometers per pixel. The acquired image data undergoes image segmentation, denoising, and three-dimensional reconstruction to obtain a three-dimensional digital model of the aerogel skeleton. Key parameters such as skeleton thickness, pore size distribution, and connectivity are extracted using image analysis software to form a structural feature dataset. Then, the extracted three-dimensional network structure data is standardized, unifying data of different dimensions to the 0-1 range, removing outliers, and supplementing missing values. Key feature parameters, including average pore size, porosity, specific surface area, skeleton density, and fractal dimension, are extracted. Principal component analysis is used for dimensionality reduction, retaining principal components that explain more than 95% of the cumulative variance. A mapping relationship was established between preparation process parameters and the characteristics of the three-dimensional network structure. Process parameters included raw material type, concentration, pH value, temperature, and time, while structural characteristics included porosity, specific surface area, and pore size distribution. Data augmentation techniques were employed to expand the training samples, including random rotation, scaling, and adding noise, increasing the original dataset by five times. The final dataset comprised at least 10,000 records, including raw material composition, process parameters, and structural characteristic parameters. 80% of the data was used for training, 10% for validation, and 10% for testing.

[0064] Specifically, the principle of this invention is as follows: The core technical principle of this invention lies in the organic combination of constructing a multi-layered fire protection mechanism and an intelligent optimization system. Firstly, aerogel, as the material basis, has a unique three-dimensional network structure containing a large number of nanoscale pores. These pores are filled with still air, effectively inhibiting heat transfer through convection and conduction. Simultaneously, the discontinuity of the aerogel skeleton also blocks the heat propagation path, which is the physical basis for achieving low thermal conductivity. The aerogel's particle size range of 1 to 100 micrometers is designed to maintain good thermal insulation performance while achieving uniform dispersion with the polyurethane system, forming a stable composite structure.

[0065] Secondly, this invention constructs a carbon monoxide catalytic conversion system by loading a horgalat agent onto the surface of an aerogel, utilizing its high specific surface area. Under high-temperature fire conditions, when organic matter thermally decomposes to produce carbon monoxide, the horgalat agent catalyzes its reaction with oxygen to generate carbon dioxide. This process follows a redox catalytic mechanism, lowering the activation energy of carbon monoxide and accelerating its conversion rate. Simultaneously, by loading flame retardants (such as ammonium polyphosphate, expanded graphite, etc.) onto the aerogel, a synergistic effect of forming a char layer and gas-phase flame retardancy upon heating further improves the material's fire resistance.

[0066] Third, the aerogel structure optimization network model established in this invention mathematically models the relationship between the complex three-dimensional network structure characteristics of aerogels and the macroscopic properties of the material. This model extracts the spatial features of the aerogel using a three-dimensional convolutional neural network, and then identifies key connection points in the structure using a graph attention network, thereby predicting the impact of the aerogel on the overall material properties. The model training data includes structural data and corresponding performance parameters of aerogel samples under different preparation conditions, and the optimal process conditions are derived in reverse through mapping relationships. In practical applications, deviation data obtained through performance testing is fed back to the model's attention mechanism to dynamically adjust feature weights, achieving online optimization of the model and ensuring precise control of process parameters.

[0067] This invention establishes a theoretical framework through flame propagation suppression kinetics equations, heat conduction barrier equations, and equations relating aerogel porosity to thermal insulation performance, closely integrating microstructure design with macroscopic performance control. The design of components A and B follows the principles of polyurethane chemical reactions, ensuring optimal crosslinking and physical properties by precisely controlling the functional group equivalent ratio of polyols and polyisocyanates. The aging, surface modification, and gradient vacuum drying processes throughout the preparation process guarantee the integrity and hydrophobicity of the aerogel structure, solving the problems of easy breakage and moisture absorption in traditional aerogels in composite materials. This allows the final product to maintain excellent thermal insulation performance while also possessing the characteristics of suppressing toxic gases and structural stability.

[0068] The following provides a specific embodiment 1 of the present invention, and the specific implementation of each step in this embodiment 1 is described in detail below.

[0069] The specific implementation of step S01 involves selecting one of silica aerogel, carbon aerogel, or alumina aerogel as a raw material and preparing an aerogel sol with a particle size of 1 to 100 micrometers using the sol-gel method. Specifically, a silicon precursor, such as tetraethyl orthosilicate or tetramethoxysilane, is first dissolved in ethanol. Deionized water is added to initiate a hydrolysis reaction. A catalyst, such as ammonia, is added to adjust the pH to 8-9. The reaction temperature is controlled at 25-30°C, and the reaction time is 4-6 hours to ensure complete hydrolysis of the silicon precursor. Then, a condensation reaction is carried out at 50-60°C for 6-8 hours to form a sol system with a preliminary network structure. The sol particle size is monitored in real-time using dynamic light scattering technology, and the reaction is stopped when it reaches the 1-100 micrometer range. The resulting sol is then treated with ultrasonic dispersion technology for 10-15 minutes to ensure uniform particle size distribution. During the sol-gel reaction, the hydrolysis reaction of the silicon precursor can be represented as follows:

[0070] Si(OR)4+4H2O→Si(OH)4+4ROH;

[0071] In the formula, Si(OR)4 is a silicon precursor, R is an alkyl group, H2O is a water molecule, Si(OH)4 is a hydrolysis product, and ROH is an alcohol byproduct.

[0072] The polycondensation reaction process can be represented as: nSi(OH)4 → (SiO2) n +2nH2O;

[0073] In the formula, n is the degree of polymerization of the condensation reaction, and the value of n ranges from 100 to 1000, (SiO2). n The resulting silicon dioxide network structure.

[0074] The growth kinetics equation for sol particle size can be expressed as:

[0075] In the formula, r is the sol particle size in micrometers; t is the reaction time in hours; k g (T) is the temperature-dependent growth rate constant, in micrometers per hour; C s r represents the concentration of reactants in the solution, expressed in moles per liter. max The maximum achievable particle size, in micrometers, r max The value ranges from 100 to 120 micrometers. The purpose of this step is to prepare an aerogel sol with a three-dimensional network structure, low density, and high porosity. This structure is the basis for subsequent loading of functional materials and achieving excellent fire resistance.

[0076] The specific implementation of step S02 involves adding hogalat agent to the aerogel sol prepared in step S01 to prepare an aerogel-supported carbon monoxide catalyst via a loading process. Specifically, firstly, the hogalat agent is prepared into an aqueous solution at a ratio of 3-5% of the aerogel mass, and the pH value of the solution is controlled within the range of 5.5-6.5 using a pH adjustment method. Then, the hogalat agent aqueous solution is mixed with the aerogel sol using an ultrasonic-assisted impregnation method. The ultrasonic power is set to 300-400 watts for 20-30 minutes to ensure that the hogalat agent is uniformly dispersed in the aerogel sol. Next, a crosslinking agent such as dimethyldimethoxysilane is added at a concentration controlled at 0.5-1.5%. After stirring evenly, the mixture is allowed to stand for 12-24 hours to allow the sol to transform into a gel state. The loading amount of hogalat agent in the aerogel can be expressed as:

[0077] In the formula, L CO Catalyst loading, in percentage; m cat The mass of hogalat is expressed in grams; m gel The value represents the mass of the aerogel, expressed in grams.

[0078] The reaction kinetic equation for the oxidation of carbon monoxide catalyzed by hogalat can be expressed as:

[0079] In the formula, [CO] is the carbon monoxide concentration, in moles per liter; t is the reaction time, in seconds; k cat [O2] is the catalytic reaction rate constant, expressed in square meters per mole per second; [O2] is the oxygen concentration, expressed in moles per liter; S cat k represents the specific surface area of ​​the catalyst, expressed in square meters per gram. cat The relationship with temperature follows the Arrhenius equation:

[0080] In the formula, A is the pre-exponential factor, with a value range of 10. 6 ~10 8 square meters per mole per second; E a The activation energy is 50–70 kJ / mol; R is the gas constant, with a value of 8.314 J / mol / Kelvin; and T is the temperature in Kelvin. The purpose of this step is to ensure that the hogalat agent is uniformly distributed within the three-dimensional network structure of the aerogel, thereby achieving highly efficient catalytic oxidation of carbon monoxide produced in a fire.

[0081] The specific implementation of step S03 involves adding the flame retardant to the aerogel sol prepared in step S01, and preparing the aerogel-loaded flame retardant through a loading process. Specifically, one or more flame retardants selected from ammonium polyphosphate, expanded graphite, glass microspheres, or montmorillonite are prepared as a suspension at a ratio of 15-20% of the aerogel mass. The flame retardant suspension is mixed with the aerogel sol using high-shear emulsification technology, with the shear rate controlled at 8000-10000 rpm for 15-20 minutes to ensure uniform dispersion of the flame retardant in the aerogel sol. The uniformity of the flame retardant dispersion in the aerogel can be evaluated by the dispersion coefficient.

[0082] In the formula, D c σ is the dispersion coefficient, dimensionless; c The standard deviation of flame retardant concentration in a local area is expressed as a percentage. D represents the average concentration of the flame retardant, expressed as a percentage. When the flame retardant is uniformly dispersed, D... c It should be less than 0.2.

[0083] The total amount of flame retardant added can be expressed as:

[0084] In the formula, m fr The total mass of the flame retardant is expressed in grams; w i The mass percentage of the i-th flame retardant, in percentage form; m gelThe value represents the mass of the aerogel, expressed in grams; n represents the type and quantity of flame retardants. The purpose of this step is to ensure that the flame retardants are evenly distributed within the three-dimensional network structure of the aerogel, enabling it to exert multiple flame-retardant effects under fire conditions, such as hindering heat conduction, absorbing heat, and forming a protective layer.

[0085] The specific implementation of step S04 involves preparing component A by uniformly mixing polyol, aerogel-supported carbon monoxide catalyst, aerogel-supported flame retardant, and foaming agent in a specific ratio. Specifically, polyether polyol is selected as the main component, with a molecular weight range of 3000–6000, a hydroxyl value of 350–450 mg / g potassium hydroxide, and a content of 60–85% by mass. 0.1–3% by mass of the aerogel-supported carbon monoxide catalyst prepared in step S02 and 10–20% by mass of the aerogel-supported flame retardant prepared in step S03 are added. A foaming agent such as dichlorofluoroethane or cyclopentane is selected, with a content of 2–10% by mass. The mass ratio of each component in component A can be expressed as:

[0086] In the formula, w A，i m is the mass percentage of the i-th component in component A; A，i Let represent the mass of the i-th component in component A, in grams; the value of i ranges from 1 to 4, corresponding to polyol, aerogel-supported carbon monoxide catalyst, aerogel-supported flame retardant, and foaming agent, respectively. The hydroxyl content of the polyol can be expressed as:

[0087] In the formula, [OH] represents the hydroxyl content, in milligrams of potassium hydroxide per gram; 56.1 represents the molar mass of potassium hydroxide, in grams per mole; f represents the hydroxyl functionality, ranging from 2 to 3; M n The value represents the number-average molecular weight of the polyol, expressed in grams per mole. The purpose of this step is to mix the polyol with hydroxyl functional groups and functional additives to form a homogeneous component A, providing a basis for subsequent reaction with component B.

[0088] The specific implementation of step S05 involves preparing component B by uniformly mixing the polyisocyanate and aerogel in a specific ratio. Specifically, polymeric MDI (diphenylmethane diisocyanate) is selected as the polyisocyanate, with an NCO content of 30-32% and a total mass percentage of 90-95%. The original aerogel prepared in step S01 is added, with a content of 5-10% by mass. The mass ratio of each component in component B can be expressed as:

[0089] In the formula, w B，j m is the mass percentage of the j-th component in component B; B，jLet be the mass of the j-th component in component B, in grams; j ranges from 1 to 2, corresponding to polyisocyanate and aerogel, respectively. The NCO content of polyisocyanate can be expressed as:

[0090] In the formula, [NCO] represents the isocyanate group content, in percentage; 42 represents the molar mass of the NCO group, in grams per mole; g represents the isocyanate functionality, ranging from 2 to 3; M w The values ​​represent the mass-average molecular weight of the polyisocyanate, expressed in grams per mole. The purpose of this step is to prepare component B containing isocyanate functional groups, where the added aerogel enhances the mechanical and fire-resistant properties of the final product.

[0091] The specific implementation of step S06 involves using an aerogel structure optimization network model to predict and analyze the three-dimensional network structure of the aerogel and determine the optimal process parameters. Specifically, the raw material composition, proportioning data, and preparation conditions from steps S01 to S05 are first input into the aerogel structure optimization network model. This model, based on deep learning principles, employs an architecture combining a three-dimensional convolutional neural network and a graph attention network, enabling it to predict the microstructural characteristics of aerogels under different conditions. The input parameter vector of the aerogel structure optimization network model can be represented as: X = [x1, x2, ..., x...]. 64 ] T ;

[0092] In the formula, X is the input parameter vector with a dimension of 64×1; x1 to x 64 These represent different input characteristics, including aerogel type, particle size distribution, porosity, density, and degree of surface modification.

[0093] Optionally, the convolution operation of a 3D convolutional neural network can be represented as: Z (l) =f(W (l) *Z (l-1) +b (l) );

[0094] In the formula, Z (l) Here, f is the feature map of the l-th layer; f is the activation function, usually the ReLU function; W (l) Here are the parameters of the convolution kernel in layer l; * indicates a 3D convolution operation; b (l) is the bias parameter of the l-th layer; the value of l ranges from 1 to 3, corresponding to the three convolutional layers.

[0095] Optionally, the attention coefficients of a graph attention network can be calculated as follows:

[0096] In the formula, α ijLet be the attention coefficient of node i to node j; a be the attention vector; W be the weight matrix; h be the attention coefficient of node i to node j; a be the attention vector of node i; W be the weight matrix of node j; h be the attention coefficient of node i to node j; a be the attention vector ... i and h j These are the feature vectors of node i and node j, respectively; || represents the vector concatenation operation. Let i represent the set of neighboring nodes of node i.

[0097] Optionally, the model's output parameter vector can be represented as: Y = [y1, y2, ..., y...]. 16 ] T ;

[0098] In the formula, Y is the output parameter vector with a dimension of 16×1; y1 to y 16 These represent different output characteristics, including process parameters such as mixing temperature, stirring rate, molding pressure, and curing time, as well as predicted performance indicators such as material density, compressive strength, thermal conductivity, and fire resistance limit. The purpose of this step is to use computer models to help determine the optimal process parameters, improve the stability and predictability of product performance, and reduce the number of experiments and material waste.

[0099] The specific implementation of step S07 involves mixing component A and component B according to the process parameters determined in step S06 to achieve foaming and molding of the material. Specifically, component A and component B are accurately weighed at a mass ratio of 1:1 to 1:1.5, and the optimal ratio is determined based on the functional group equivalent ratio. The functional group equivalent ratio between component A and component B can be expressed as:

[0100] In the formula, r is the functional group equivalent ratio; [OH] is the hydroxyl content of the polyol in component A, in moles per gram; m A The total mass of component A is expressed in grams; w A，1 [NCO] represents the mass fraction of polyols in component A; [NCO] represents the isocyanate group content of polyisocyanates in component B, in moles per gram; m B The total mass of component B is expressed in grams; w B，1 This represents the mass fraction of polyisocyanate in component B. To ensure optimal curing results, the value of r should range from 0.9 to 1.1.

[0101] The reaction kinetics between isocyanate groups and hydroxyl groups during the mixed reaction can be expressed as:

[0102] In the formula, [NCO] is the isocyanate group concentration, in moles per liter; t is the reaction time, in seconds; k r is the reaction rate constant, in liters per mole per second; [OH] is the hydroxyl concentration, in moles per liter. rThe relationship with temperature follows the Arrhenius equation:

[0103] In the formula, A r The pre-exponential factor has a value range of 10. 3 ~10 5 Liters per mole per second; E ar The activation energy is 40–60 kJ / mol; R is the gas constant, with a value of 8.314 J / mol / Kelvin; and T is the temperature in Kelvin.

[0104] The flame propagation suppression kinetic equation following the foaming process can be expressed as:

[0105] In the formula, V f The flame propagation rate is expressed in millimeters per second; t is time, expressed in seconds; k f C is the flame propagation rate constant, measured in millimeters per second; fr This refers to the concentration of flame retardant, expressed as a percentage by mass; C fr，max [O2] represents the maximum effective concentration of the flame retardant, expressed as a mass percentage; [O2] represents the oxygen concentration, expressed as a volume percentage; K O ρ is the oxygen concentration half-saturation constant, expressed as a volume percentage; ρ is the material density, expressed as kilograms per cubic meter; ρ0 is the reference density, expressed as kilograms per cubic meter.

[0106] The thermal conduction barrier equation can be expressed as: λ eff =λ s ·(1-φ)+λ g ·φ+λ r ;

[0107] In the formula, λ eff λ is the effective thermal conductivity of the material, measured in watts per meter (Kelvin). s φ represents the thermal conductivity of the solid framework, measured in watts per meter Kelvin; φ represents the porosity of the material, dimensionless; λ g λ is the thermal conductivity of a gas, measured in watts per meter Kelvin; r Contribution to radiative heat conduction, measured in watts per meter Kelvin.

[0108] The equation relating aerogel porosity to thermal insulation performance can be expressed as:

[0109] In the formula, λ eff λ is the overall thermal conductivity of the material, measured in watts per meter Kelvin; sλ is the thermal conductivity of the skeleton material, measured in watts per meter Kelvin; φ is the porosity, dimensionless; n is an empirical parameter, typically taken as 1.5–2.0; λ g σ is the gas thermal conductivity coefficient, measured in watts per meter Kelvin; β is the gas thermal conductivity adjustment coefficient, dimensionless; Kn is the Knudsen number, dimensionless; σ is the Stefan-Boltzmann constant, with a value of 5.67 × 10⁻⁶. -8 Watts per square meter per Kelvin to the fourth power; T is temperature, in Kelvin; ρ b ρ represents the bulk density of the material, in kilograms per cubic meter; e represents the extinction coefficient of the material, in meters. The purpose of this step is to form a fire-resistant sealing material with a closed-cell structure by controlling the polyurethane chemical reaction between components A and B, combined with the properties of aerogel.

[0110] The specific implementation of step S08 involves conducting performance tests on the prepared aerogel two-component foamed fireproof sealing material to ensure it meets relevant technical specifications. Specifically, the material density is first tested according to GB / T6343 standard, using Archimedes' principle. The sample size is 100×100×100 mm, and the measurement temperature is 23±2℃. The required density is controlled between 120 and 180 kg / m³. The compressive strength is then tested according to GB / T8813 standard, using a sample size of 50×50×50 mm, a compression rate of 5 mm / min, and a test temperature of 23±2℃. The required compressive strength is not less than 1.0 MPa. Finally, the thermal conductivity is tested according to GB / T10295 standard, using the heat flow meter method. The sample size is 300×300×25 mm, and the test temperature is 25℃. The required thermal conductivity is not higher than 0.040 W / m Kelvin. The fire resistance limit is tested according to GB / T9978 standard. The sample size is determined based on the actual application scenario, typically 1000×1000×100 mm, and the required fire resistance limit is not less than 2 hours. The performance deviation can be calculated as follows:

[0111] In the formula, ΔP i P represents the deviation of the i-th performance indicator, expressed as a percentage. i，mea P is the measured value of the i-th performance index; i，tar Let be the target value for the i-th performance index; the value of i ranges from 1 to 4, corresponding to density, compressive strength, thermal conductivity, and fire resistance limit, respectively. The purpose of this step is to comprehensively evaluate the physical and mechanical properties and fire resistance of the material, providing data support for subsequent optimization.

[0112] The specific implementation of step S09 involves calculating the performance deviation weights using the performance deviation data obtained in step S08, and optimizing the aerogel structure optimization network model. Specifically, the deviations of each performance index are first normalized, converting deviation data of different dimensions into a unified scale. The normalization process can be expressed as:

[0113] In the formula, ΔP is the normalized deviation of the i-th performance index, dimensionless; i Let be the original deviation of the i-th performance index; min(ΔP) is the minimum deviation of all performance indexes; max(ΔP) is the maximum deviation of all performance indexes.

[0114] The relative importance weights of each performance indicator are determined using the analytic hierarchy process (AHP), and a judgment matrix is ​​constructed.

[0115] In the formula, a ij This represents the importance ratio of performance metric i to performance metric j. a ii =1. For fire-resistant sealing materials, usually a 12 =2, a 14 =1,

[0116] Calculate the largest eigenvalue λ of the judgment matrix max and the corresponding eigenvector w: Aw=λ max w;

[0117] In the formula, λ max To determine the largest eigenvalue of matrix A; w = [w1, w2, w3, w4] T The corresponding feature vectors represent the weights of each performance index. Typically, w1 = 0.2, w2 = 0.1, w3 = 0.3, and w4 = 0.4, corresponding to the weights of density, compressive strength, thermal conductivity, and fire resistance limit, respectively.

[0118] Calculate the overall performance deviation weight:

[0119] In the formula, W P The weighting for overall performance deviation is dimensionless; w i Let be the weight of the i-th performance metric; Let be the normalized deviation of the i-th performance index.

[0120] The calculated performance deviation weights are input into the attention mechanism of the aerogel structure optimization network model to update the attention weights:

[0121] In the formula, The updated attention coefficient; α ij γ is the original attention coefficient; γ is the learning rate adjustment parameter, ranging from 0.1 to 0.5; W PThis is used to weight the overall performance deviation. The purpose of this step is to optimize and update the theoretical model online using actual production data, so that the model's predictions better reflect actual production conditions.

[0122] The specific implementation of step S10 involves re-predicting the optimal process parameters based on the aerogel structure optimization network model optimized in step S09, and then performing quality control. In practice, the current production conditions and raw material parameters are first input, the optimized aerogel structure optimization network model is run, and the new process parameter prediction results are obtained. The new process parameters can be expressed as: P new =f opt (X, θ) new );

[0123] In the formula, P new For the newly predicted process parameter vector; f opt X is the optimized model function; X is the input parameter vector; θ new This is the optimized set of model parameters.

[0124] The adjustment amount of the process parameter can be expressed as: ΔP = P new -P old ;

[0125] In the formula, ΔP is the vector of process parameter adjustments; P new P represents the newly predicted vector of process parameters. old This is the original process parameter vector.

[0126] Adjust the actual process parameters according to the adjustment amount: P adj =P old +α·ΔP;

[0127] In the formula, P adj This represents the adjusted process parameter vector; α is the parameter adjustment coefficient, ranging from 0.6 to 0.9, used to avoid over-adjustment. Specific adjustment ranges include: mixing temperature ±2℃, stirring rate ±200 rpm, molding pressure ±0.05 MPa, and curing time ±4 hours.

[0128] Implement the adjusted process parameters for small-batch trial production, test product performance, and calculate new performance deviations:

[0129] In the formula, The deviation of the new i-th performance metric is expressed as a percentage. The measured value of the i-th performance index after process adjustment; P i，tar Let be the target value of the i-th performance metric.

[0130] Computational performance improvement rate:

[0131] In the formula, I P Performance improvement rate, in percentage; w i The weight of the i-th performance metric; |ΔP i | represents the absolute value of the deviation from the original performance index; This represents the absolute value of the deviation from the new performance metric. When I P When I > 10%, it indicates that the process adjustment is effective; when I P When the percentage is less than 5%, the model parameters need to be re-optimized. The purpose of this step is to continuously improve product performance and manufacturing processes through closed-loop feedback optimization, ensuring stable product performance and meeting fire-stopping requirements.

[0132] The detailed implementation of the aerogel structure optimization network model involves constructing a deep learning model combining multi-layer convolutional neural networks and graph neural networks. This model comprises four main parts: an input layer, a feature extraction layer, a structure prediction layer, and an output layer. The input layer receives parameters such as raw material composition and process conditions, and is a 64×1 vector. The mathematical representation of the input layer is: X = [x1, x2, ..., x...]. 64 ] T ;

[0133] In the formula, X is the input feature vector; x i This represents the i-th feature parameter, where i ranges from 1 to 64. Specifically, x1 to x5 represent aerogel type parameters, x6 to x... 15 Represents the particle size distribution parameter, x 16 ~x 25 x represents the porosity parameter. 26 ~x 35 Denotes the density parameter, x 36 ~x 45 The parameter representing the degree of surface modification, x 46 ~x 64 This indicates process condition parameters.

[0134] The feature extraction layer employs a three-dimensional convolutional neural network, consisting of three convolutional blocks. The operation of the first convolutional block can be represented as: Z (1) =ReLU(Conv3D(X) reshape W (1) b (1) ));

[0135] In the formula, Z (1) The output feature map of the first convolutional layer has dimensions of 32×16×16×16; ReLU is the modified linear unit activation function; Conv3D is the three-dimensional convolution operation; X reshape To reshape the input vector X into a four-dimensional tensor with dimensions 1×16×16×16; W(1) b represents the kernel parameters of the first convolutional layer, which contains 32 3×3×3 convolutional kernels; (1) represents the bias parameters of the first convolutional layer, with a dimension of 32×1.

[0136] The operation of the second convolutional block can be represented as: Z (2) =ReLU(Conv3D(Z) (1) W (2) b (2) ));

[0137] In the formula, Z (2) The output feature map of the second convolutional layer has dimensions of 64×8×8×8; W (2) b represents the kernel parameters for the second convolutional layer, which contains 64 3×3×3 convolutional kernels; (2) represents the bias parameters for the second convolutional layer, with a dimension of 64×1.

[0138] The operation of the third convolutional block can be represented as: Z (3) =ReLU(Conv3D(Z) (2) W (3) b (3) ));

[0139] In the formula, Z (3) The output feature map of the third convolutional layer has dimensions of 128×8×8×8; W (3) b represents the kernel parameters for the third convolutional layer, which contains 128 3×3×3 convolutional kernels; (3) is the bias parameter for the third convolutional layer, with a dimension of 128×1.

[0140] The structure prediction layer uses a graph attention network, which first converts the feature map into graph structure data: G = (V, E);

[0141] In the formula, G represents the graph structure; V represents the set of nodes, where each node represents an intersection point in the aerogel framework; and E represents the set of edges, representing framework connections. The node characteristics in the graph can be represented as: H = [h1, h2, ..., h...]. N ] T ;

[0142] In the formula, H is the node feature matrix; h i Let N be the feature vector of the i-th node, with a dimension of 128×1; N is the number of nodes, usually 512.

[0143] The first layer operation of a graph attention network can be represented as:

[0144] In the formula, H (1)The output features of the attention layer in the first image are shown; || represents the concatenation operation; K is the number of attention heads, with a value of 8; σ is the activation function, using ELU. Represents the set of neighboring nodes of node i; The attention coefficients of node i to node j are calculated for the k-th attention head; W k Let be the weight matrix for the k-th attention head, with dimensions 64×128. The attention coefficient is calculated using the following formula:

[0145] In the formula, a k is the attention vector of the k-th attention head, with a dimension of 128×1; LeakyReLU is the modified linear unit activation function with leakage, and the slope of the negative half axis is 0.2.

[0146] Similarly, the operation of the attention layer in the second graph is represented as follows:

[0147] In the formula, H (2) The output features of the attention layer in the second graph; The attention coefficient calculated for the k-th attention head in the second layer; W k(2) This is the weight matrix for the k-th attention head in the second layer, with a dimension of 64×64; These are the node features output from the first layer.

[0148] The third attention layer uses a single attention head, and the operation is represented as follows:

[0149] In the formula, H (3) The output features of the third graph's attention layer; The attention coefficients calculated for the third layer; W (3) This is the weight matrix for the third layer, with dimensions 64×64; The node features are the output of the second layer. The attention coefficients of the third layer are normalized using the Softmax function.

[0150] In the formula, a (3) This is the attention vector for the third layer, with a dimension of 128×1.

[0151] The output layer maps the output features of the graph attention network to the target parameter space through a fully connected network:

[0152]

[0153] In the formula, Z fc1 The output of the first fully connected layer has a dimension of 256×1; W fc1 This is the weight matrix of the first fully connected layer, with dimensions of 256×(N×64); b is the vector flattened from the output features of the third attention layer; fc1 The bias parameters for the first fully connected layer are 256×1.

[0154] Apply Dropout to prevent overfitting:

[0155] In the formula, The features are those after applying Dropout; p = 0.3 means that the values ​​are randomly set to zero with a probability of 0.3.

[0156] Second fully connected layer operations:

[0157] In the formula, Z fc2 This is the output of the second fully connected layer, with a dimension of 128×1; W fc2 b is the weight matrix of the second fully connected layer, with dimensions of 128×256; fc2 The bias parameters for the second fully connected layer are 128×1.

[0158] Final output layer operation: Y = W out Z fc2 +b out ;

[0159] In the formula, Y is the final output of the model, with a dimension of 16×1; W out b is the weight matrix of the output layer, with dimensions 16×128; out The output layer contains bias parameters with a dimension of 16×1. The output includes: y1~y4 representing process parameters (mixing temperature, stirring rate, molding pressure, curing time); y5~y8 representing aerogel structural parameters (average pore size, porosity, skeleton density, specific surface area); and y9~y... 16 This indicates the predicted material properties (density, compressive strength, thermal conductivity, fire resistance limit, etc.).

[0160] The detailed steps for establishing the training dataset for the aerogel structure optimization network model involve systematically collecting and processing aerogel structure and performance data. First, three-dimensional network structure data of aerogel samples under different preparation conditions are collected. Microscopic morphology images of the aerogels are obtained using techniques such as scanning electron microscopy, transmission electron microscopy, and X-ray tomography, with an image resolution of at least 10 nanometers per pixel. The equation for acquiring the aerogel microstructure images can be expressed as: I(x, y, z) = f img (M, R, C, I0);

[0161] In the formula, I(x, y, z) is the image intensity at the three-dimensional spatial location (x, y, z); f imgdenoted as the imaging function; M is the material property parameter; R is the imaging device resolution parameter; C is the imaging condition parameter; and I0 is the incident light intensity.

[0162] The acquired image data is segmented using a thresholding method:

[0163]

[0164] In the formula, S(x, y, z) is the segmented binary image; T is the segmentation threshold, which is usually automatically determined by the Otsu method and ranges from 20% to 40% of the image grayscale value.

[0165] The segmented 2D slices are reconstructed into a 3D model using a 3D reconstruction algorithm:

[0166]

[0167] In the formula, V 3D For the reconstructed 3D voxel model; R 3D For 3D reconstruction algorithm; S i (x, y) is the two-dimensional segmented image of the i-th layer; n z The number of slices is usually 200 to 500.

[0168] Extract structural parameters of the aerogel framework, such as the framework thickness distribution:

[0169] In the formula, D sk (r) is the probability density function for a skeleton with thickness r; N(r) is the number of voxels in the skeleton with thickness r; N total This represents the total number of skeleton voxels.

[0170] Calculation of pore size distribution:

[0171] In the formula, D pore V(d) is the probability density function for pore size d; V(d) is the pore volume for diameter d; V total This represents the total pore volume.

[0172] Aerogel connectivity calculation:

[0173] In the formula, C is the connectivity parameter, which is dimensionless; V max V is the volume of the most connected component; total This represents the total pore volume.

[0174] The extracted structural parameters are standardized.

[0175] In the formula, The standardized parameter value; xi The original parameter value; μ i σ is the mean of this parameter; i This represents the standard deviation of the parameter.

[0176] Dimensionality reduction using principal component analysis: X PCA =X norm W PCA ;

[0177] In the formula, X PCA X is the eigenvalue matrix after dimensionality reduction; norm W is the standardized original feature matrix. PCA The total number of master network layers. Parameter initialization uses the He initialization method: The weight matrix for the principal component analysis is composed of eigenvectors. The choice of the number of principal components k satisfies:

[0178] In the formula, λ i Let be the i-th feature value; n is the original feature dimension.

[0179] Data augmentation techniques include random rotation transformations: X rot =R(θ, φ, ψ)X;

[0180] In the formula, X rot The data is rotated; R(θ, φ, ψ) is a three-dimensional rotation matrix, where θ, φ, and ψ are Euler angles, randomly taking values ​​in the range [0, 2π].

[0181] Random scaling transformation: X scale =sX;

[0182] In the formula, X scale is the scaled data; s is the scaling factor, which randomly takes a value in the range [0.9, 1.1].

[0183] Add Gaussian noise:

[0184] In the formula, x noise Data after adding noise; With a mean of 0 and a variance of Gaussian noise, σ noise The value ranges from 5% to 10% of the standard deviation of the original data.

[0185] The final training dataset contains no fewer than 10,000 samples, with 80% used for training, 10% for validation, and 10% for testing. The dataset structure can be represented as follows:

[0186] In the formula, For the complete dataset; X iY is the input feature vector of the i-th sample; i Let N be the output label vector of the i-th sample; N is the total number of samples, not less than 10000.

[0187] The training process of the aerogel structure optimization network model first initializes the network parameters:

[0188] In the formula, θ represents the set of all parameters of the model; W (l) and b (l) These are the weights and bias parameters of the l-th layer, respectively; L is the network...

[0189] In the formula, Indicates a normal distribution; n in This is the input dimension for this layer.

[0190] The loss function is defined as a combination of mean squared error and regularization term:

[0191]

[0192] In the formula, Here is the loss function; N batch For batch size, typically 32 to 64; Y i f(X) represents the true label of the i-th sample. i ;θ) is the model's predicted output for the i-th sample; λ is the regularization coefficient, with a value of 10. -4 ~10 -5 .

[0193] Parameters are updated using the Adam optimization algorithm:

[0194] In the formula, θ t and θ t+1 These represent the parameters at step t and step t+1, respectively; η is the learning rate, initially set to 0.001. and These are the correction estimates for the first and second moments, respectively; ∈ is the stability constant, with a value of 10. -8 .

[0195] Use a learning rate scheduling strategy:

[0196] In the formula, η t Let be the learning rate at step t; η0 be the initial learning rate; γ be the decay rate, with a value of 0.9; T decay The decay period is usually set to the training set size divided by the batch size.

[0197] An early stopping strategy is employed to prevent overfitting; when the validation set loss is continuous N... patience Training is stopped if there is no improvement in the next iteration.

[0198] In the formula, N is the validation set loss for the t-th iteration; patience For patience parameters, a value of 5 to 10 is typically used.

[0199] The online optimization and update process, which fine-tunes the model based on actual test results, employs the gradient descent method.

[0200]

[0201] In the formula, θ new and θ old These represent the model parameters before and after fine-tuning, respectively; η fine To fine-tune the learning rate, it is usually set to 0.1 times the initial learning rate; This represents the gradient of the loss function with respect to the parameters, calculated on the new data.

[0202] To better understand and implement this invention, Example 2 of a specific application scenario is provided below: Researchers prepared a high-performance fireproof sealing material for cable penetrations in subway tunnels using the method of this invention. First, silica aerogel sol was prepared by mixing 120 kg of tetraethyl orthosilicate with 200 kg of anhydrous ethanol, adjusting the pH to 8.5 with ammonia, and adding 50 kg of deionized water. The mixture was then hydrolyzed at 28°C for 6 hours. Next, the mixture was heated to 55°C for a condensation reaction for 7.5 hours. Dynamic light scattering technology was used to monitor the reaction, and the reaction was stopped when the sol particle size reached 45 μm, yielding 100 portions of silica aerogel sol.

[0203] 2.5 parts of hogalat agent were prepared into an aqueous solution, and the pH was adjusted to 6.0. This solution was then mixed with the aerogel sol using an ultrasonic-assisted impregnation method (350 W ultrasonic power, 25 minutes). 12 parts of dimethyldimethoxysilane were added as a crosslinking agent. After thorough stirring, the mixture was allowed to stand at room temperature for 18 hours to allow the sol to transform into a gel. The gel was aged at 65°C for 60 hours, and then surface-modified using trimethylchlorosilane vapor deposition at 85°C for 7 hours. Finally, the aerogel-supported hogalat catalyst was dried under gradient reduced pressure: the initial pressure was atmospheric pressure, decreasing to 3 kPa at a rate of 12 kPa per hour, while maintaining the temperature at 55°C for 30 hours, yielding the catalyst for later use.

[0204] Simultaneously, 100 parts of silica aerogel sol were prepared. 12 parts of expanded graphite and 8 parts of ammonium polyphosphate were mixed at a mass ratio of 3:2 and then mixed with the aerogel sol using a high-shear emulsification technique (shear rate 9500 rpm, time 18 minutes). 11 parts of tetraethoxysilane were added as a crosslinking agent. After stirring evenly, the mixture was allowed to stand for 22 hours to form a gel. The gel was aged at 70℃ for 66 hours, and then surface modified with hexyltrimethoxysilane (concentration 4%) at 80℃ for 8.5 hours via liquid-phase impregnation. Finally, the mixture was dried under gradient reduced pressure, starting at atmospheric pressure and decreasing to 4 kPa at a rate of 10 kPa per hour, while maintaining the temperature at 60℃ for 36 hours, yielding the aerogel-loaded flame retardant for later use.

[0205] Researchers mixed 78 parts of polyether polyol with a molecular weight of 4500 and a hydroxyl value of 400 mg KOH / g, 7 parts of cyclopentane blowing agent, 13 parts of aerogel-supported flame retardant, and 2 parts of aerogel-supported catalyst at 1350 rpm for 18 minutes, with the temperature controlled at 23°C. After mixing, the mixture was degassed at 70 kPa for 12 minutes to obtain component A. Simultaneously, component B was prepared by mixing 97 parts of polymeric MDI with an NCO content of 31.5% and 3 parts of alumina aerogel at 700 rpm for 12 minutes, with the temperature controlled at 18°C, and then degassed at 75 kPa for 10 minutes.

[0206] Based on the predictions of the aerogel structure optimization network model, the optimal process parameters were determined as shown in Table 1.

[0207] Table 1 Optimal process parameters for aerogel two-component foamed fireproof sealing material

[0208] Process parameters numerical values Mixing temperature 25℃ Stirring rate 2000 rpm Molding pressure 0.18MPa Curing time 28 hours Mass ratio of components A / B 1:1.2 Isocyanate index 1.08

[0209] The researchers mixed component A and component B at a mass ratio of 1:1.2 according to the process parameters in Table 1, stirred at 2000 rpm for 13 seconds at 25°C, poured the mixture into a mold pre-coated with a release agent, maintained it under a pressure of 0.18 MPa, and demolded it after curing at room temperature (23±2°C) and relative humidity of 50% for 28 hours.

[0210] Researchers analyzed the microscopic morphology images of the obtained aerogels, such as... Figure 2The images shown are microscopic images of aerogels prepared using process parameters determined based on the staff's experience and the optimal process parameters. The red circles in the two sub-images represent the small grids formed by the aerogels. (a) is a microscopic image of the aerogels prepared using process parameters determined based on the staff's experience; (b) is a microscopic image of the aerogels prepared using the optimal process parameters. By comparison, it can be seen that the skeleton formed by the edge of each small grid of aerogel in (b) is more delicate, which can effectively increase the volume of the small grids of the aerogel and is beneficial for better catalyst loading.

[0211] The prepared materials were subjected to performance tests, and the results are shown in Table 2.

[0212] Table 2 Performance test results of aerogel two-component foamed fireproof sealing material

[0213] Performance indicators Target value Measured value Deviation rate (%) <![CDATA[Density (kg / m 3 )]]> 150 147 -2.0 Compressive strength (MPa) 1.2 1.25 +4.2 Thermal conductivity (W / (m·K)) 0.038 0.036 -5.3 Fire resistance rating (h) 3.0 3.2 +6.7 Anti-aging properties (strength retention rate after 1000h) ≥90 94.5 +5.0 CO conversion rate (%) ≥95 97.8 +2.9

[0214] Researchers applied the prepared aerogel two-component foamed fireproof sealing material to the cable penetration point of a subway tunnel and compared the performance differences with those of traditional organic fireproof sealing materials (polyurethane-based). The results are shown in Table 3.

[0215] Table 3 Comparison of the properties of the material of this invention with those of traditional materials

[0216]

[0217] In practical applications, this fire-resistant sealing material, after being installed in subway tunnels, underwent high-temperature combustion tests (according to GB / T 31802 standard). The material demonstrated excellent fire resistance, significant heat insulation, and effectively catalytically decomposed carbon monoxide produced during fires, reducing smoke toxicity. It also possesses excellent mechanical properties and aging resistance; random inspections three years after installation showed stable performance with no obvious signs of aging.

[0218] Traditional fire-stopping materials primarily employ organic polymers (such as polyurethane and epoxy resin) mixed with inorganic flame-retardant fillers. These materials suffer from drawbacks such as high density, low fire resistance, high thermal conductivity, and the production of toxic gases during combustion. This invention combines aerogel technology with two-component foaming technology, utilizing the ultra-low thermal conductivity of the aerogel's three-dimensional network structure, combined with a loaded hogallat catalyst and flame retardant, to significantly improve the material's overall performance. Compared to traditional technologies, this invention reduces material density by 33.2%, thermal conductivity by 40.0%, increases fire resistance by 113.3%, and reduces smoke toxicity by 78.7%. Particularly regarding smoke toxicity, the hogallat catalyst effectively catalyzes the conversion of CO produced in a fire into CO2, greatly reducing the risk of fire-related injuries and fatalities. Simultaneously, the application of an optimized aerogel structure network model allows for more precise optimization of process parameters, resulting in more stable and reliable product performance and enabling large-scale industrial production.

[0219] It should be noted that the variables involved in this invention are explained in detail in Tables 4, 5, 6, 7, and 8 below.

[0220] Table 4. Variable Explanation Table (Part 1)

[0221]

[0222]

[0223] Table 5. Variable Explanation Table (Part Two)

[0224]

[0225] Table 6. Variable Explanation Table (Part 3)

[0226]

[0227]

[0228] Table 7. Variable Explanation Table (Part Four)

[0229]

[0230] Table 8. Variable Explanation Table (Part 5)

[0231]

[0232]

[0233] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for preparing aerogel two-component foamed fireproof plugging material, characterized in that, The application relates to a preparation method of a carbon monoxide catalyst and a fire retardant loaded aerogel. The application comprises the following steps: preparing an aerogel sol; loading a carbon monoxide catalyst and a fire retardant into the aerogel sol; preparing an A component containing the carbon monoxide catalyst and the fire retardant loaded aerogel and a B component containing the aerogel; determining optimal process parameters by using an aerogel structure optimization network model; mixing the A component and the B component according to the predicted process parameters to form a foam by a reaction, and the foaming process follows a flame propagation inhibition dynamics equation, a heat conduction barrier equation and an aerogel porosity and heat insulation performance relationship equation; testing the performance of the product and obtaining performance deviation data; and optimizing the aerogel structure optimization network model by using the performance deviation data and re-determining the process parameters. The step of preparing the aerogel sol is specifically characterized by selecting one of silica aerogel, carbon aerogel or alumina aerogel as a raw material, and preparing the aerogel sol with a particle size of 1-100 microns by using a sol-gel method, wherein the aerogel has a three-dimensional network structure and is low in density and high in porosity. The step of loading the carbon monoxide catalyst is specifically characterized by uniformly dispersing a Hoggart agent into the aerogel sol, adding a crosslinking agent to make the sol form a gel state, and performing aging, surface modification and gradient pressure reduction drying treatment to obtain the carbon monoxide catalyst loaded aerogel. The step of loading the fire retardant is specifically characterized by uniformly dispersing one or more of ammonium polyphosphate, expanded graphite, glass microspheres or montmorillonite into the aerogel sol, adding a crosslinking agent to make the sol form a gel state, and performing aging, surface modification and gradient pressure reduction drying treatment to obtain the fire retardant loaded aerogel.

2. The method of claim 1, wherein, The step of preparing the A component is specifically characterized by uniformly mixing 60-85% of a polyol, 0.1-3% of the carbon monoxide catalyst loaded aerogel, 10-20% of the fire retardant loaded aerogel and 2-10% of a foaming agent to obtain the A component, wherein the foaming agent is one of monofluorodichloroethane or cyclopentane.

3. The method of claim 2, wherein, The step of preparing the B component is specifically characterized by uniformly mixing 90-95% of a polyisocyanate and 5-10% of the aerogel according to a mass ratio to obtain the B component.

4. The method of claim 3, wherein, The mass ratio of the A component to the B component is 1:1-1:1.5, and the mass ratio ensures that the material reaches the optimal crosslinking degree and physical performance in the curing reaction.

5. The method of claim 4, wherein, The performance deviation data refers to the difference between the actual performance index and the expected target performance index, including density deviation, compressive strength deviation, thermal conductivity deviation and fire resistance limit deviation.

6. The method of claim 5, wherein, The specific structure of the aerogel structure optimization network model is a deep learning model combining a multi-layer convolutional neural network and a graph neural network, and the model comprises an input layer, a feature extraction layer, a structure prediction layer and an output layer; the feature extraction layer adopts a three-dimensional convolutional neural network to extract the spatial features of the aerogel material, the structure prediction layer adopts a graph attention network to capture key connection points in the skeleton structure of the aerogel, and the output layer predicts the three-dimensional network structure characteristics of the aerogel and the influence of the three-dimensional network structure characteristics on the performance of the material.

7. The method of claim 6, wherein, ​ 8. The method of claim 7, wherein, ​ 9. The method of claim 8, wherein, ​ 10. The method of claim 9, wherein, The optimization step of the aerogel structure optimization network model is specifically that a performance deviation weight is calculated by using the performance deviation data, the performance deviation weight is input into an attention mechanism of the aerogel structure optimization network model, the attention weight distribution in the aerogel structure optimization network model is adjusted, and the aerogel structure optimization network model is updated online.

Images