Fireproof coating preparation method based on multi-dimensional digital twinning

Abstract

This invention relates to the field of fire-retardant coating preparation technology, and more particularly to a method for preparing fire-retardant coatings based on multi-dimensional digital twins. This method uses bamboo sap, a byproduct of bamboo carbonization, as the core base material. First, it collects environmental parameters of the target wood structure application scenario, physical performance parameters of the wood structure substrate, and coating performance data to construct a multi-dimensional digital twin model coupling multiple physical fields of the environment, substrate, and coating. Through transient simulation, it performs performance prediction on a large number of candidate formulations, and optimizes the formulation with multi-constraints, focusing on maximizing the fire resistance limit. The coating is prepared according to the optimized formulation, and a closed-loop iteration is formed through performance verification and model correction, ultimately obtaining a fire-retardant coating suitable for the target scenario. This invention significantly improves coating R&D efficiency, realizes high-value utilization of bamboo processing byproducts, ensures precise adaptation of the coating to the application scenario, and combines environmental benefits with engineering practicality.

Description

Technical Field

[0001] This invention relates to the field of fire-retardant coating preparation technology, specifically a method for preparing fire-retardant coatings based on multi-dimensional digital twins. Background Technology

[0002] With the rapid development of green building and prefabricated building industries, timber-framed buildings, with their numerous advantages such as low carbon footprint, environmental friendliness, thermal insulation, excellent seismic performance, and short construction cycles, are seeing continuous expansion in their application across various fields, including civil buildings, cultural and tourism buildings, and ancient building restoration. Wood is a combustible material, prone to pyrolysis and combustion in fire scenarios, leading to a rapid decline in structural load-bearing capacity and potentially causing building collapses and other safety accidents. Therefore, fire protection is a core issue that must be addressed in the design and application of timber-framed buildings. Fire-retardant coatings, applied to the surface of the timber substrate, can form a dense, heat-insulating protective layer through expansion and carbonization during a fire, slowing down the pyrolysis and combustion process of the wood. This is currently the most widely used and effective fire protection method for timber-framed buildings.

[0003] Most mainstream fire-retardant coatings for wood structures currently on the market use petroleum-based synthetic resins as film-forming materials. Their production process is highly dependent on non-renewable chemical raw materials. Furthermore, some solvent-based coatings have high emissions of volatile organic compounds, which can adversely affect the environment and human health during production, construction, and service, thus being incompatible with the green and low-carbon attributes of wood structures. Existing fire-retardant coatings are mostly general-purpose formulations, unable to be specifically optimized for the environmental characteristics of different application scenarios and the thermophysical properties of different types of wood structural substrates. This results in significant fluctuations in fire protection effectiveness during actual engineering applications, easily leading to insufficient protection or performance redundancy, making it difficult to achieve precise matching of coating performance with application scenarios.

[0004] The traditional development model for fire-retardant coatings relies primarily on the experience of researchers to conduct formula trial and error. Formula optimization is achieved through extensive physical formulation mixing and repeated fire resistance testing. This process suffers from significant drawbacks, including long development cycles, high material consumption, and high trial-and-error costs. Furthermore, the traditional model struggles to achieve synergistic optimization of multi-dimensional performance indicators. While improving the coating's fire resistance limit, it often leads to problems such as uncontrolled coating expansion, decreased adhesion, and deteriorated application performance, failing to balance fire resistance, mechanical properties, and application performance. More importantly, the traditional model cannot accurately simulate the real-world service environment and fire thermal exposure process of the coating in the target application scenario. It cannot fully reproduce the multi-physics coupling effects between the environmental field, the coating, and the wood-based substrate. This results in significant discrepancies between the actual protective performance of laboratory-developed formulas and laboratory test results in practical engineering applications, failing to meet the real-world fire protection requirements of the target scenario.

[0005] Digital twin technology, through digital modeling and multiphysics coupled simulation, can accurately simulate and predict the behavior of physical objects throughout their entire lifecycle. While it has already achieved large-scale application in several industrial manufacturing sectors, its application in the research and development of fire-retardant coatings remains significantly lacking. Existing technologies can only simulate simple thermophysical parameters of a single coating, failing to construct multi-dimensional coupled digital twin models covering the environmental field, the thermal response of the wood structure substrate, and the dynamic fire-retardant performance of the coating. This prevents a complete simulation of the full-process thermal response behavior of coated wood structures under fire scenarios, and the simulation prediction accuracy is insufficient to support precise optimization of coating formulations. Furthermore, existing technologies lack a closed-loop iterative mechanism for simulation prediction and physical verification, making it impossible to continuously optimize model accuracy using measured data or achieve bidirectional optimization and improvement of coating formulations and digital twin models. This hinders the efficient and precise research and development of fire-retardant coatings.

[0006] In summary, existing fire-retardant coating technologies for wood structures cannot simultaneously address a series of industry pain points, such as insufficient environmental performance, low utilization efficiency of biomass byproducts, poor compatibility between formulations and application scenarios, low R&D efficiency, and difficulty in synergistic optimization of multiple performance aspects. There is an urgent need to develop a novel method for preparing fire-retardant coatings and to construct a closed-loop R&D system covering the entire process from scenario data collection, simulation prediction, formulation optimization, preparation and production to performance verification. This system would not only enable high-value utilization of bamboo sap water byproducts but also significantly improve the R&D efficiency, environmental performance, and scenario compatibility of fire-retardant coatings, providing high-performance, low-cost, green, and low-carbon fire protection solutions for wood-structured buildings. Summary of the Invention

[0007] The purpose of this invention is to provide a method for preparing fire-retardant coatings based on multi-dimensional digital twins, so as to solve the problems mentioned in the background art.

[0008] To achieve the above objectives, the present invention provides the following technical solution: A method for preparing fire-retardant coatings based on multi-dimensional digital twins includes the following steps: S1. Collect time-series data of environmental parameters of the target wood structure application scenario, physical performance parameters of wood structure substrate, and performance parameter database of fireproof coatings based on bamboo sap, a by-product of bamboo carbonization. S2. Construct an environmental field digital twin sub-model based on environmental parameter time-series data; construct a wood structure thermodynamic response digital twin sub-model based on the physical performance parameters of the wood structure substrate and according to the heat conduction equation and pyrolysis reaction kinetic equation; construct a coating performance digital twin sub-model based on the performance parameter database using a data-driven method; load the environmental boundary conditions output by the environmental field digital twin sub-model onto the wood structure thermodynamic response digital twin sub-model, and use the coating dynamic thermophysical parameters output by the coating performance digital twin sub-model as the surface boundary conditions of the wood structure thermodynamic response digital twin sub-model, thereby realizing the coupling of multiple physical fields of environment, substrate, and coating, and establishing a multi-dimensional digital twin model; S3. Input multiple candidate bamboo sap water-based fireproof coating formulations into a multi-dimensional digital twin model and perform multi-physics field coupled transient simulation to simulate the temperature field evolution, carbonization layer formation and growth process of the wood structure under preset fire scenarios or heat exposure conditions. Output the performance prediction results of the fire resistance limit, coating expansion behavior and substrate pyrolysis depth change with time for each candidate formulation. S4. Based on the preset optimization objectives, an optimization algorithm is used to find the best coating formulation among multiple candidate formulations and select the preferred coating formulation that meets the optimization objectives. The optimization objectives include maximizing the fire resistance limit, ensuring that the coating expansion ratio is within the set range, and ensuring that the coating adhesion meets the standard. S5. According to the preferred coating formula, bamboo sap is mixed with film-forming aid, dehydration catalyst, charring agent, foaming agent, filler and additives in proportion, and then dispersed and ground to obtain fireproof coating. S6. The performance of the prepared fireproof coating is verified. If the deviation between the verification result and the performance prediction result of the corresponding formula exceeds the preset threshold, the parameters of the multi-dimensional digital twin model are corrected according to the verification result. Steps S3 to S5 are repeated based on the corrected model until the deviation between the verification result and the performance prediction result is within the preset threshold, and the final fireproof coating adapted to the target wood structure application scenario is obtained.

[0009] As a preferred option, the environmental parameter time series data should include at least temperature, humidity and solar radiation intensity; the physical performance parameters of the wood structure substrate should include at least density, specific heat capacity, thermal conductivity and pyrolysis kinetic parameters; the performance parameter database should include the thermal conductivity, specific heat capacity, expansion ratio, char formation rate and initial expansion temperature of the fireproof coating under different formulations.

[0010] As can be seen from the technical solution provided by the present invention above, the fire-retardant coating preparation method based on multi-dimensional digital twins provided by the present invention has the following beneficial effects: This invention significantly improves the efficiency of fire-retardant coating R&D and reduces R&D and trial-and-error costs. Traditional fire-retardant coating R&D relies on a large number of repeated physical formulation experiments and performance tests, resulting in long R&D cycles, high material consumption, and high costs. This invention, through multi-physics field coupled transient simulation of a multi-dimensional digital twin model, can quickly complete the performance prediction and virtual screening of a large number of candidate formulations, significantly reducing the number of formulations and the frequency of physical experiments, effectively shortening the R&D cycle, reducing material and labor costs in the R&D process, and improving the flexibility and response speed of formulation R&D. This invention achieves precise matching between coating formulations and target application scenarios, solving the industry pain point of the disconnect between formulation design and actual service environment. By collecting real environmental parameters and substrate physical performance parameters of the target wood structure application scenario, this invention constructs a multi-dimensional coupled digital twin model covering environmental field, substrate thermal response, and coating performance. It can accurately simulate the full-process thermal response behavior of wood structures coated with paint under real fire scenarios, so that the formulation design is fully aligned with the fire protection requirements and service conditions of the target scenario. This avoids the problem of insufficient or redundant protection performance of general formulations in specific scenarios, and greatly improves the pertinence and reliability of fire protection for wood structures. This invention realizes the high-value resource utilization of agricultural and forestry processing by-products, combining environmental benefits with synergistic performance advantages. Using bamboo sap, a by-product generated during bamboo carbonization, as the core base material for fire-retardant coatings, this invention effectively solves the environmental pollution problem caused by the indiscriminate discharge of bamboo processing by-products, achieving waste recycling. Simultaneously, it reduces the coating's dependence on petroleum-based synthetic resins and other chemical raw materials, minimizing the environmental impact during coating production and use. The polyphenolic organic char-forming components contained in bamboo sap can synergistically enhance the fire-retardant system of the coating, increasing the char formation rate and density of the carbonized layer after heating, further strengthening the coating's heat insulation and fire-retardant performance, thus achieving a dual improvement in environmental benefits and fire-retardant performance. This invention constructs a multi-objective collaborative formulation optimization system to achieve a balance and improvement in the overall performance of coatings. With maximizing fire resistance limit as the core optimization objective, this invention also uses key service performance characteristics such as coating expansion ratio and adhesion as rigid constraints. Through hierarchical screening and intelligent optimization algorithms, it completes the targeted optimization of formulations, effectively resolving the contradiction between improving single performance and balancing overall performance in traditional formulation development. This ensures that the final selected formulation possesses excellent fire resistance and heat insulation performance, stable physical and mechanical properties, and good construction compatibility, meeting all performance requirements for fire-retardant coatings for wood structures. This invention establishes a closed-loop iterative mechanism for simulation prediction and physical verification, achieving bidirectional optimization and improvement of digital models and coating performance. Through measured data from physical performance verification of coatings, this invention performs sensitivity analysis and inversion calibration on key parameters of the multi-dimensional digital twin model, continuously improving the model's simulation prediction accuracy. Simultaneously, based on the corrected model, it re-initiates formulation optimization and preparation verification, forming a closed-loop optimization system throughout the entire process. This mechanism ensures a high degree of match between the actual service performance of the coating and the simulation design expectations, effectively reducing the risk of performance deviations in engineering applications. Furthermore, the continuously improving digital twin model provides a mature simulation foundation for subsequent coating development in similar scenarios, possessing strong technical reusability and scalability. This invention establishes a standardized and replicable coating preparation process, suitable for industrial production and large-scale engineering applications. Targeting the characteristics of bamboo sap water-based fire-retardant coatings, this invention constructs a standardized preparation process covering the entire process from raw material purification and pretreatment, graded dispersion, fine grinding to finished product performance control. It clarifies the key process parameters and quality control requirements for each stage, effectively ensuring the performance uniformity and stability of different batches of coatings, thus meeting the needs of industrial mass production. The resulting bamboo sap water-based fire-retardant coating is compatible with the construction techniques of conventional wooden structures and can be directly applied to fire protection projects for various types of wooden structures, providing a high-performance, low-cost, and environmentally friendly standardized solution for fire safety in wooden structures. Attached Figure Description

[0011] Figure 1 This is a schematic diagram of the steps in the preparation method of fire-retardant coating based on multi-dimensional digital twins according to the present invention. Detailed Implementation

[0012] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0013] To better understand the above technical solutions, the following will provide a detailed explanation of the technical solutions in conjunction with the accompanying drawings and specific embodiments.

[0014] like Figure 1 As shown, this embodiment of the invention provides a method for preparing fire-retardant coatings based on multi-dimensional digital twins, including the following steps: S1. Collect time-series data of environmental parameters of the target wood structure application scenario, physical performance parameters of wood structure substrate, and performance parameter database of fireproof coatings based on bamboo sap, a by-product of bamboo carbonization. S2. Construct an environmental field digital twin sub-model based on environmental parameter time-series data; construct a wood structure thermodynamic response digital twin sub-model based on the physical performance parameters of the wood structure substrate and according to the heat conduction equation and pyrolysis reaction kinetic equation; construct a coating performance digital twin sub-model based on the performance parameter database using a data-driven method; load the environmental boundary conditions output by the environmental field digital twin sub-model onto the wood structure thermodynamic response digital twin sub-model, and use the coating dynamic thermophysical parameters output by the coating performance digital twin sub-model as the surface boundary conditions of the wood structure thermodynamic response digital twin sub-model, thereby realizing the coupling of multiple physical fields of environment, substrate, and coating, and establishing a multi-dimensional digital twin model; S3. Input multiple candidate bamboo sap water-based fireproof coating formulations into a multi-dimensional digital twin model and perform multi-physics field coupled transient simulation to simulate the temperature field evolution, carbonization layer formation and growth process of the wood structure under preset fire scenarios or heat exposure conditions. Output the performance prediction results of the fire resistance limit, coating expansion behavior and substrate pyrolysis depth change with time for each candidate formulation. S4. Based on the preset optimization objectives, an optimization algorithm is used to find the best coating formulation among multiple candidate formulations and select the preferred coating formulation that meets the optimization objectives. The optimization objectives include maximizing the fire resistance limit, ensuring that the coating expansion ratio is within the set range, and ensuring that the coating adhesion meets the standard. S5. According to the preferred coating formula, bamboo sap is mixed with film-forming aid, dehydration catalyst, charring agent, foaming agent, filler and additives in proportion, and then dispersed and ground to obtain fireproof coating. S6. The performance of the prepared fireproof coating is verified. If the deviation between the verification result and the performance prediction result of the corresponding formula exceeds the preset threshold, the parameters of the multi-dimensional digital twin model are corrected according to the verification result. Steps S3 to S5 are repeated based on the corrected model until the deviation between the verification result and the performance prediction result is within the preset threshold, and the final fireproof coating adapted to the target wood structure application scenario is obtained.

[0015] In this embodiment, the environmental parameter time series data includes at least temperature, humidity and solar radiation intensity; the physical performance parameters of the wood structure substrate include at least density, specific heat capacity, thermal conductivity and pyrolysis kinetic parameters; the performance parameter database contains the thermal conductivity, specific heat capacity, expansion ratio, char formation rate and initial expansion temperature of the fireproof coating under different formulations; The core function of step S1 is to complete the full collection and standardized processing of basic data required for the construction of a multi-dimensional digital twin model. This covers three core dimensions: environmental characteristics of the target wood structure application scenario, thermophysical properties of the wood structure substrate, and performance of bamboo sap water-based fireproof coating. This provides complete and reliable data support for subsequent sub-model construction, multi-physics coupling simulation, and formula optimization. The detailed steps are as follows: Step S1-1: Time-series data acquisition and standardization processing of environmental parameters for the target timber structure application scenario: Multiple environmental sensor nodes are deployed in typical locations of the target wood structure application scenario. Typical locations include the sunny side of the wood structure, the shady side, the enclosed indoor space, the area around the ventilation opening, and areas with high fire risk. Each sensor node integrates three types of acquisition units: temperature, humidity, and solar radiation intensity, to achieve synchronous real-time acquisition of environmental parameters. The data collection process needs to simulate the environmental change patterns of the target scenario throughout the entire cycle, with continuous data collection for no less than 30 calendar days and a data collection interval of no more than 10 minutes, to ensure coverage of environmental parameter fluctuation characteristics under different weather conditions and at different times. The collected raw environmental parameter time series data are preprocessed. The preprocessing process includes three steps: outlier removal, missing value imputation, and data standardization. Outlier removal is achieved using the 3σ criterion. For time-series datasets with a single environmental parameter, the mean and standard deviation of the dataset are calculated, and outlier data points exceeding the range of the mean plus or minus three standard deviations are removed. The corresponding calculation formula is as follows: ,in, For the time series dataset, the first The original data values ​​at each acquisition time. This is the arithmetic mean of the time series dataset. This represents the standard deviation of the time series dataset. Missing value imputation is achieved using linear interpolation. For missing data segments that occur within no more than three consecutive acquisition cycles after outlier removal, the imputation value at the missing position is calculated through linear fitting based on the adjacent valid data points before and after the missing segment. The corresponding calculation formula is: ,in, For the first Missing value imputation results at each acquisition time. The nearest valid data value preceding the missing segment. The nearest valid data value after the missing segment. for The corresponding data collection time, for The corresponding data collection time, For the first The timestamp corresponding to each collection moment; Data standardization employs a min-max normalization method, mapping all environmental parameter data, after outlier removal and missing value imputation, to a uniform numerical range of 0 to 1. This eliminates dimensional differences between parameters, resulting in a standardized time-series dataset of environmental parameters. The corresponding calculation formula is: ,in, For the standardized results of a single set of data, The original valid data values ​​to be standardized. This is the minimum value in the time series dataset for this type of parameter. The maximum value in the time series dataset for this type of parameter; The standardized time-series data of temperature, humidity, and solar radiation intensity are aligned according to the time of collection and stored in the environmental parameter time-series dataset to provide input data for the construction of the digital twin sub-model of the environmental field. Step S1-2: Collection of physical performance parameters and construction of parameter set for wood structure substrate: Core samples were drilled from the same batch of wood structure substrates in the target wood structure application scenario. The number of core samples was no less than 5 sets, and the size and specifications of each set of core samples were uniform to ensure the representativeness and repeatability of the test results. For each core sample, density testing, specific heat capacity testing, thermal conductivity testing, and pyrolysis kinetic parameter testing were carried out sequentially to obtain the core physical performance parameters of the substrate. The density test was performed using the water displacement method. The density of the substrate was calculated by measuring the dry weight of the core sample and the volume of water displaced. The specific heat capacity and thermal conductivity were tested using a thermal constant analyzer. The test environment temperature ranged from 25 degrees Celsius to 600 degrees Celsius, with a test temperature interval of no more than 50 degrees Celsius. Data on the changes in specific heat capacity and thermal conductivity of the substrate at different temperatures were obtained, and the corresponding relationship between the thermal properties of the substrate and temperature was established. Thermogravimetric analysis (TGA) was used to test the pyrolysis kinetic parameters. The testing process employed a nitrogen atmosphere to simulate an oxygen-deficient fire environment. Three heating rates were set: 5°C / min, 10°C / min, and 20°C / min. The test temperature range was from room temperature to 800°C. Thermogravimetric curves and mass change rate curves of the substrate as a function of temperature were recorded. Based on the obtained thermogravimetric data, three core pyrolysis kinetic parameters—pyrolysis activation energy, pre-exponential factor, and reaction order—were calculated using the isoconversion method. The corresponding calculation formulas were constructed based on Arrhenius's law, and the expression is: ,in, The pyrolysis conversion rate of the substrate. For the pyrolysis reaction rate, Pre-exponential factor, This is the pyrolysis reaction mechanism function. The activation energy for pyrolysis, This is the universal gas constant. Thermodynamic temperature; All the substrate density, specific heat capacity and thermal conductivity at different temperatures, and pyrolysis kinetic parameters obtained from the tests were summarized. Abnormal test groups with a dispersion coefficient greater than 5% were removed, and the arithmetic mean of the remaining valid test groups was taken as the final parameter result. A set of physical performance parameters of wood structure substrate was established to provide input data for the construction of a digital twin model of the thermal response of wood structure. Step S1-3: Construction of the performance parameter database for bamboo sap-based fireproof coatings: Using bamboo sap, a byproduct of bamboo carbonization, as the base material, multiple candidate coating formulations with different component ratios were designed. Each formulation contains seven core components: bamboo sap, film-forming aid, dehydration catalyst, charring agent, foaming agent, filler, and additives. The types of components in each formulation are kept consistent, and only the mass ratio of a single component is adjusted to form a candidate formulation matrix covering the entire ratio range. According to the mixing ratio requirements of each candidate formulation, the coating samples were prepared in sequence. The preparation process includes three core steps: raw material mixing, high-speed dispersion, and grinding and filtration, to ensure that the preparation process of all samples is consistent and to eliminate the interference of process differences on the performance test results. For each group of prepared coating samples, three types of core performance tests were carried out in sequence: thermogravimetric analysis, thermal conductivity test, and refractory chamber test, to obtain five types of core performance parameters of different formulation coatings: thermal conductivity, specific heat capacity, expansion ratio, char formation rate, and initial expansion temperature. Thermogravimetric analysis (TGA) was performed using a thermogravimetric analyzer. The testing process simulated a real fire environment with oxygen by using an air atmosphere. The heating rate was set to 10 degrees Celsius per minute, and the test temperature range was from room temperature to 1000 degrees Celsius. The thermogravimetric curve of sample mass change with temperature was recorded. Based on the test data, the char formation rate and initial pyrolysis temperature of the sample were calculated. At the same time, the specific heat capacity and thermal conductivity of the sample at different temperatures were obtained. The thermal conductivity was tested using the hot wire method, with the test temperature range from 25 degrees Celsius to 800 degrees Celsius and the test temperature interval not exceeding 100 degrees Celsius. The dynamic thermal conductivity change data of the coating during the entire process of thermal expansion was obtained, and the corresponding relationship between thermal conductivity and temperature was established. The fire resistance chamber test was carried out in accordance with the standard fire resistance test specifications. The coating sample was uniformly coated on the substrate test board according to the standard coating thickness. After the coating was completely cured, it was placed in the fire resistance test furnace and the heat exposure test was carried out according to the standard temperature rise curve. The data of coating thickness change with temperature and the data of unexposed surface temperature change with time were recorded in real time. After the test, the final expansion thickness of the coating was measured and the maximum expansion ratio and the initial expansion temperature of the coating were calculated. The component ratio information, test results of five core performance parameters, and thermophysical property parameter change curves of all candidate formulations were correlated. Sample data with abnormalities such as coating peeling and cracking during the test were removed. After normalizing the valid test data, a performance parameter database of bamboo sap water-based fireproof coating was constructed, providing complete sample data for the training and verification of the digital twin sub-model of coating performance.

[0016] In this embodiment, the core function of step S2 is to construct digital twin sub-models of the environmental field, the thermal response of the wood structure, and the coating performance based on the comprehensive basic data collected in step S1. Through the directional loading and dynamic linkage of multi-physics boundary conditions, deep coupling of the environment, substrate, and coating is achieved. Ultimately, a multi-dimensional digital twin model is established that can accurately simulate the comprehensive thermal response behavior of a wood structure coated with fire-retardant paint under real environmental and fire conditions. This provides a core simulation carrier and computational foundation for subsequent candidate formulation simulation calculations, performance prediction, and targeted optimization. The detailed steps are as follows: Step S2-1: Construction of the digital twin model of the environmental field: The standardized environmental parameter time-series dataset generated in step S1 is used as the model input, and a digital twin sub-model of the environmental field is constructed using time series modeling methods. The modeling process uses the acquisition time as the time dimension benchmark and the sensor node deployment location as the spatial dimension benchmark to fit and model the spatiotemporal distribution characteristics of three types of parameters: temperature, humidity, and solar radiation intensity. This achieves an accurate description of the spatiotemporal distribution patterns and dynamic trends of temperature, humidity, and radiation fields in the target wood structure application scenario. During the model training process, the standardized environmental parameter time-series dataset is divided into a training set and a validation set in a 7:3 ratio. The training objective is to minimize the mean square error between the predicted and measured values ​​of the time-series data, thereby completing the iterative optimization of the model parameters. After the model training is completed, it can output the real-time environmental parameters at the corresponding location and the time-varying thermal environmental parameters under the preset fire scenario based on the input time and spatial location information, providing environmental boundary conditions for subsequent multi-physics coupling. Step S2-2: Construction of a digital twin model of the thermal response of the timber structure: Using the set of physical performance parameters of the wood structure substrate established in step S1 as input, a heat conduction equation describing the evolution of the internal temperature field of the wood structure is established based on Fourier's law of heat conduction, and a pyrolysis reaction kinetic equation describing the pyrolysis reaction rate of the wood structure material is established based on Arrhenius's law. The two equations are bidirectionally coupled to construct a digital twin model of the thermodynamic response of the wood structure. This model can accurately simulate the dynamic evolution of the internal temperature distribution and pyrolysis degree of the wood structure under heating conditions. The heat conduction equation adopts the three-dimensional unsteady-state heat conduction governing equation, and its expression is: ,in, The density of the wood structure substrate. The specific heat capacity of the wood structure substrate. Thermodynamic temperature For time, The thermal conductivity of the wood structure substrate is given. This is the internal heat source term generated by the pyrolysis reaction of the wood structure; The pyrolysis reaction kinetic equation is constructed based on Arrhenius's law and is used to describe the change in substrate conversion rate with temperature and time during pyrolysis. The expression is as follows: ,in, The pyrolysis conversion rate of the substrate. Pre-exponential factor, The activation energy for pyrolysis, This is the universal gas constant. This is a pyrolysis reaction mechanism function; During the model construction process, the pyrolysis conversion rate and internal heat source terms calculated by the pyrolysis reaction kinetic equation are actually substituted into the heat conduction equation. At the same time, the temperature field distribution data output by the heat conduction equation is fed back to the pyrolysis reaction kinetic equation in real time, realizing the two equations bidirectional coupling. After the coupling is completed, the model can output the temperature field distribution, pyrolysis conversion rate distribution and carbonization layer development status inside the wood structure in real time according to the input environmental boundary conditions and surface coating boundary conditions. Step S2-3: Construction of a digital twin model of coating performance: The performance parameter database of bamboo sap water-based fireproof coating constructed in step S1 is used as a training sample. A deep learning algorithm is used to train the model and establish a digital twin sub-model of coating performance. This model can realize end-to-end mapping from coating formulation parameters to dynamic performance parameters of the coating throughout the heating process, and accurately output the corresponding coating thermal conductivity, specific heat capacity, expansion ratio and char rate change curves throughout the heating process. Before model training, the coating performance parameter database was divided into training, validation, and test sets in an 8:1:1 ratio. The training set was used for iterative updates of the model weight parameters, the validation set was used for model performance monitoring and overfitting prevention during training, and the test set was used to verify the model's generalization ability after training. The model input consisted of the mass percentage parameters of each component in the coating formulation, and the output consisted of continuous curves showing the changes in thermal conductivity, specific heat capacity, expansion ratio, and char rate of the coating with temperature within the temperature range of room temperature to 1000 degrees Celsius. The model training aimed to minimize the mean square error between the predicted and measured curves. An adaptive moment estimation algorithm was used for iterative updates of the weight parameters. After each iteration, the model performance was evaluated using the validation set. If the error decrease in the validation set was less than 0.05% for 15 consecutive iterations, the model was considered to have converged, and training was stopped. After training, the model's generalization ability was verified using the test set to ensure that the model still had stable prediction accuracy for new formulation parameters that had not been used in the training. Step S2-4: Loading and coupling of environmental field boundary conditions: The environmental boundary conditions calculated by the digital twin sub-model of the environmental field are used as external loads and applied to the uncoated surface of the digital twin sub-model of the thermal response of the wood structure. The environmental boundary conditions include three core parameters: air temperature, convective heat transfer coefficient, and radiative heat flux. All three parameters are time-varying parameters that change dynamically with time. During the loading process, the time-varying environmental parameters output by the digital twin sub-model of the environmental field are mapped point by point to the nodes of the uncoated surface of the digital twin sub-model of the thermal response of the wood structure according to the time step and spatial location, serving as the third type of boundary conditions for model solving. Through this loading method, the dynamic changes of the environmental field are accurately simulated to achieve the heat input of the uncoated surface of the wood structure, completing the first layer of coupling between the environmental field and the thermal response field of the wood structure. Step S2-5: Applying boundary conditions to coating performance and fully coupling multiphysics fields: The dynamic thermophysical properties of the coating output from the digital twin model of coating performance are used as surface boundary conditions and applied to the coated surface of the digital twin model of wood structure thermal response. The dynamic thermophysical properties of the coating include thermal conductivity, specific heat capacity and equivalent thermal resistance, which change in real time with the coating temperature and pyrolysis degree. The parameters are updated in real time with the time step of the simulation process. During the loading process, the dynamic thermophysical properties of the coating output from the digital twin model of coating performance in each time step are mapped point by point to the coated surface nodes of the digital twin model of wood structure thermal response, as the surface thermal resistance boundary conditions for model solution, so as to accurately simulate the thermal insulation and protection effect of the coating on the wood structure substrate. By applying both environmental boundary conditions and coating performance boundary conditions, a multi-physics coupling is achieved among the environmental field, coating performance, and thermal response of the wood structure, ultimately establishing a multi-dimensional digital twin model. This model can fully simulate the comprehensive thermal response behavior of a wood structure coated with fire-retardant paint under real environmental changes and preset fire heat exposure conditions, providing a complete computational framework for the performance simulation and prediction of subsequent candidate formulations.

[0017] In this embodiment, the core function of step S3 is to input multiple candidate bamboo sap water-based fire-retardant coating formulations into a multi-dimensional digital twin model that has completed multi-physics coupling, and to conduct a full-process multi-physics coupling transient simulation. This simulation fully simulates the entire process of temperature field evolution, char layer formation and growth of wood structures coated with the corresponding formulations under preset fire scenarios or heat exposure conditions. It accurately outputs the performance prediction results of each candidate formulation, including fire resistance limit, coating expansion behavior, and substrate pyrolysis depth changes over time, providing comprehensive quantitative performance data support for subsequent formulation optimization. The detailed steps are as follows: Step S3-1: Candidate Recipe Input and Recipe Parameter Set Construction: Multiple candidate bamboo sap-based fire-retardant coating formulations were input into a multi-dimensional digital twin model to construct a corresponding formulation parameter set for each candidate formulation. The formulation parameter set included the mass percentage data of seven core components: bamboo sap, film-forming aid, dehydration catalyst, charring agent, foaming agent, filler, and additives. The dimensional settings and sorting rules of the parameter set were fully matched with the input variable requirements of the coating performance digital twin sub-model. The constructed formulation parameter sets were used as input variables of the coating performance digital twin sub-model to complete the initialization configuration of the formulation parameters before simulation, ensuring that each set of candidate formulations could independently trigger the full-process simulation calculation of the model. Step S3-2: Setting and loading initial boundary conditions for simulation: A unified preset fire scenario or thermal exposure condition is set as the initial boundary condition for simulation. This condition is loaded into the digital twin sub-model of the environmental field to generate the corresponding time-varying thermal environment parameters. The preset fire scenario or thermal exposure condition can be a standard temperature rise curve or a real fire scenario temperature rise curve corresponding to the target wood structure application scenario. The curve needs to clearly show the temperature change relationship with time, as well as the initial values ​​and dynamic change laws of the corresponding convective heat transfer coefficient and radiative heat flux. During the loading process, the preset temperature rise curve and thermal parameters are discretized into time-by-time boundary condition values ​​according to the simulation time step and input into the digital twin sub-model of the environmental field. The model outputs the time-varying thermal environment parameters for the entire simulation cycle based on the input conditions, providing a unified external thermal load reference for subsequent transient simulations. Step S3-3: Multiphysics Coupled Transient Simulation and Temperature Field Calculation: A multi-dimensional digital twin model is run, and multi-physics coupled transient simulation is carried out based on the generated time-varying thermal environment parameters and the corresponding formulation parameter sets of each candidate formulation. The simulation process adopts a fixed time step, which is set to no more than 0.1 seconds, and the total simulation time is no less than the minimum fire resistance limit required by the target application scenario. Within each simulation time step, the model sequentially completes three core steps: updating environmental field parameters, calculating dynamic thermal property parameters of the coating, and calculating the internal heat conduction and pyrolysis reaction of the wood structure. The model calculates and records the global temperature field distribution data of the wood structure at different simulation times in real time. The global temperature field distribution data includes the node-by-node temperature values ​​of the coated surface, internal cross sections and unexposed surface of the wood structure, as well as the coating temperature and pyrolysis degree data at the corresponding time. The iterative update formula for the temperature of the wooden structure nodes during the simulation is as follows: ,in, for Temperature of the wooden structure joints at all times. for Temperature of the wooden structure joints at all times. This represents the simulation time step; the meanings of the other symbols remain the same as before. Step S3-4: Simulation of the formation and growth process of the carbonized layer: Based on the global temperature field distribution data of the wood structure output at each simulation moment, the digital twin model of the wood structure's thermal response is used to identify and track the movement trajectory of the isothermal surfaces on and inside the wood structure that reach the pyrolysis temperature threshold, simulating the formation and growth process of the carbonized layer; the pyrolysis temperature threshold is set as the starting temperature of the pyrolysis reaction of the wood structure substrate, which comes from the set of physical performance parameters of the wood structure substrate collected in step S1; during the simulation, the set of nodes where the internal temperature of the wood structure reaches the pyrolysis temperature threshold is marked at each moment, and the isothermal surface at the leading edge of the carbonized layer at that moment is fitted, and the spatial position of the isothermal surface relative to the original surface of the wood structure is recorded; by the positional change of the isothermal surface at the leading edge of the carbonized layer at each moment in the entire simulation cycle, the dynamic process of the carbonized layer from its formation to its continuous growth is completely restored; The real-time calculation formula for the pyrolysis depth of the substrate is: ,in, for The depth of substrate pyrolysis at any given time. for The spatial position of the isothermal surface at the forefront of the carbonized layer at any given time. The reference spatial position of the original surface of the wooden structure; Step S3-5: Extraction of key performance indicators: From the simulation results of the entire simulation cycle, three types of core key performance indicators were extracted for each candidate formulation. The first type is the fire resistance limit, which is the simulation time required for the average temperature of the unexposed surface of the wood structure to reach the critical value. The critical value is set to 220 degrees Celsius according to the corresponding fire resistance test standard. The second type is the coating expansion behavior, which is the dynamic curve of the coating thickness changing with temperature during the entire simulation cycle as a quantitative representation of the coating expansion behavior. The curve includes the real-time coating thickness, maximum expansion thickness and corresponding temperature at each moment. The third type is the substrate pyrolysis depth, which is the data of the change in the position of the carbonized layer front relative to the original surface of the wood structure over time during the entire simulation cycle as a quantitative representation of the substrate pyrolysis depth. The data includes the real-time pyrolysis depth and maximum pyrolysis depth at each moment. The formula for calculating the maximum expansion ratio of the coating is: ,in, This represents the maximum expansion ratio of the coating. This represents the maximum expansion thickness of the coating after heating. This refers to the initial coating thickness. Step S3-6: Generation and associated storage of performance prediction result sets: The three types of core performance indicators extracted were subjected to time-series alignment and feature quantification. Time-series alignment was based on the simulation time step, unifying the time dimensions of the three types of data—fire resistance limit, coating expansion behavior, and substrate pyrolysis depth—to the same simulation time axis to ensure that the time nodes of different indicators corresponded completely. Feature quantification was performed on the time-series data of each indicator to standardize the values ​​and extract feature values, generating a performance prediction result set that included the changes of fire resistance limit, coating expansion behavior, and substrate pyrolysis depth over time. The generated performance prediction result set was uniquely associated with the corresponding candidate formulations and stored in a structured manner, providing complete quantitative data basis for formulation optimization in subsequent steps.

[0018] In this embodiment, the core function of step S4 is to build upon the full set of candidate formulation performance prediction results output in step S3. Based on the fire protection design requirements, construction and service specifications, and coating preparation feasibility of the target wood structure application scenario, a multi-dimensional, multi-constraint optimization system is constructed. Formulas that do not meet the basic performance requirements are eliminated through hierarchical, progressive constraint screening. Then, a targeted iterative optimization is conducted within the effective formulation set using an intelligent optimization algorithm. Finally, the optimal coating formulation that simultaneously meets the requirements of optimal fire resistance, suitable expansion characteristics, and satisfactory adhesion performance is selected. This provides a precise, feasible, and target-scenario-adaptive component ratio basis for the subsequent industrial-scale preparation of the coating, achieving a precise connection between simulation prediction and practical application of the bamboo sap water-based fire-retardant coating formulation. The detailed steps are as follows: Step S4-1: Construction and Standardization Preprocessing of the Candidate Recipe Comprehensive Dataset First, from the full performance prediction results set corresponding to each candidate formulation output in step S3, extract the predicted value of fire resistance limit, the full-cycle expansion behavior curve of the coating, the predicted value of maximum expansion ratio, the predicted value of coating initial expansion temperature, the full-cycle change curve of substrate pyrolysis depth, and the predicted value of maximum pyrolysis depth for each formulation; simultaneously retrieve the measured values ​​of coating adhesion, measured values ​​of coating charring rate, and all parameters of component mass ratio corresponding to each candidate formulation during the construction of the coating performance parameter database in step S1; associate and bind all the above data with the unique identifier of the corresponding candidate formulation to form a single formulation full-dimensional data unit; All single-formulation data units across all dimensions are subjected to integrity verification, and invalid data units with missing core data or misaligned data time sequences are removed. Outlier removal is performed on the numerical data in the remaining valid data units. The 3σ criterion is used to identify and remove outlier data with excessive dispersion to ensure the stability and reliability of the dataset. After outlier removal, all numerical indicators are subjected to min-max normalization to eliminate the dimensional differences between different indicators. Finally, a comprehensive candidate formulation dataset containing formulation component parameters, simulation prediction performance data, and basic performance measured data is constructed. The formula for normalization is: ,in, This is the normalization result for a single set of data. The original valid data values ​​to be processed. This is the minimum value of the indicator in the full valid dataset. This represents the maximum value of the metric in the entire valid dataset; Step S4-2: Construction of Multi-Objective Optimization System and Delineation of Optimization Space: Based on a comprehensive dataset of candidate formulations, and considering the fire protection design specifications, construction and service requirements, and coating preparation feasibility for the target timber structure application scenario, a multi-objective, multi-constraint optimization system is constructed. The optimization system is divided into three categories: core optimization objectives, hard constraints, and auxiliary verification indicators. The core optimization objective is to maximize the fire resistance limit of the coating, which directly determines the duration of safe protection for the timber structure in a fire scenario. The hard constraints consist of three items: the coating expansion ratio is within a preset adaptation range, the coating adhesion reaches a preset standard, and the initial expansion temperature of the coating is lower than the pyrolysis initiation temperature of the timber substrate. The auxiliary verification indicators are the coating charring rate and the maximum pyrolysis depth of the substrate, used for secondary rationality verification of the optimization results. Based on the above optimization system, a corresponding mathematical model and optimization space are constructed; the mathematical expression of the core optimization objective is: ,in, To optimize the objective function, The predicted fire resistance limit value for the candidate formulation; The mathematical expressions for the three hardening constraints are as follows: in, This is the lower threshold of the maximum expansion ratio of the coating. This represents the predicted maximum expansion ratio for the candidate formulation. This is the upper limit threshold for the maximum expansion ratio of the coating; ,in, These are the measured values ​​of coating adhesion corresponding to the candidate formulations. The critical value corresponding to the preset grade standard for coating adhesion is set; ,in, This is the predicted value for the initial expansion temperature of the coating. This refers to the pyrolysis initiation temperature of the wood structure substrate. The boundary of the optimization space is jointly determined by the parameter range of the candidate formulation comprehensive dataset and the prepareable range of formulation components. Each data node in the optimization space corresponds to a complete candidate formulation, ensuring that the optimization process is always carried out within the effective and feasible formulation range. Step S4-3: Fully Constrained Screening of Expansion Properties and Generation of the First Screening Recipe Set: The two constraints related to the coating expansion characteristics are transformed into a progressive numerical screening of the performance data of candidate formulations. First, the initial expansion temperature constraint screening is carried out. The predicted value of the coating initial expansion temperature of each formulation in the comprehensive dataset of candidate formulations is compared with the pyrolysis initiation temperature of the wood structure substrate. Candidate formulations with coating initial expansion temperature higher than the pyrolysis initiation temperature of the substrate are eliminated to ensure that the coating can form an expansion insulation layer before the substrate pyrolysis occurs, so as to play a fire protection role in advance. Based on the initial expansion temperature screening, the maximum expansion ratio range constraint screening is carried out. The preset lower and upper limits of the expansion ratio are determined according to the fire protection requirements of the target scenario and the service stability requirements of the coating. The lower limit is used to ensure that the coating can form a sufficiently thick dense carbonized layer after being heated to achieve effective heat insulation protection. The upper limit is used to avoid the carbonized layer structure becoming loose, the mechanical strength decreasing, and the coating becoming prone to cracking and peeling due to excessive expansion ratio, thus losing its continuous protective capability. The predicted maximum expansion ratio of the remaining candidate formulations is compared with the preset range one by one, and candidate formulations with the predicted maximum expansion ratio lower than the lower limit or higher than the upper limit are eliminated. After completing two rounds of progressive screening, the full expansion behavior curves of the remaining candidate formulations are verified a second time to ensure that the coating expansion process matches the timing of the fire temperature rise curve and that there are no abnormalities such as premature expansion failure or delayed expansion. After verification, all candidate formulations that meet the full constraints of expansion characteristics are integrated to form the first set of screening formulations. At the same time, the integrity of the formulation data in the set is verified a second time to ensure that all formulations have complete component parameters and performance data. Step S4-4: Adhesion constraint screening and generation of the second screening formula set: The constraints for achieving coating adhesion standards are transformed into a graded screening of measured adhesion data. The preset coating adhesion grade standards are determined based on the national standards corresponding to fire-retardant coatings for wood structures and the long-term service requirements of the target scenario. The measured adhesion values ​​are obtained by using the cross-cut test to obtain the grade results. The testing process is completely consistent with the process conditions for testing coating performance parameters in step S1. For the first set of formulations, the adhesion test data for each formulation is preprocessed. If there are multiple parallel test samples for the same formulation, the arithmetic mean of all valid test results is taken as the final adhesion test value. Abnormal test groups with a dispersion coefficient greater than 5% are removed to ensure the accuracy and representativeness of the adhesion data. After preprocessing, the final adhesion test value of each formulation is compared with the critical value corresponding to the preset grade standard. Candidate formulations whose adhesion test values ​​do not meet the preset grade standard are removed. After screening, the remaining candidate formulations are subjected to a second verification with all constraints to confirm that all formulations simultaneously meet the initial expansion temperature constraint, the maximum expansion ratio range constraint, and the adhesion level constraint. After verification, the remaining candidate formulations are integrated to form a second screening formulation set. All formulations in this set meet all hard constraints, providing an effective candidate population for subsequent iterative optimization. Step S4-5: Iterative optimization using intelligent optimization algorithms to determine the optimal coating formulation: A non-dominated sorting genetic algorithm was used to iteratively optimize the candidate formulations in the second screening formulation set. The optimization process took maximizing the fire resistance limit as the only core optimization objective. Three hard constraints were kept in effect throughout the process to ensure that the optimization results always met the basic performance requirements. Before the algorithm is executed, the initial configuration is completed and the core running parameters of the algorithm are set. The population size is set to the total number of recipes in the second selection recipe set, the maximum number of iterations is set to 200, the crossover probability is set to 0.7, the mutation probability is set to 0.05, and the convergence threshold is set to 0.05%. All candidate recipes in the second selection recipe set are used as the initial population, and a unique individual code is assigned to each recipe. The code content corresponds to the total component mass percentage parameter of the recipe. The algorithm iterative process is executed cyclically according to the following steps: The first step is fitness value calculation. Based on the core optimization objective function, a fitness calculation function is constructed. The higher the predicted fire resistance limit of the formulation, the higher the corresponding fitness value. For individuals that violate the hard constraints, a penalty function mechanism is introduced to reduce their fitness value, ensuring the rigidity of the constraints. The expression of the fitness calculation function is as follows: ,in, This represents the individual fitness value. The coefficients of the penalty function are... For the first The amount of violation of a constraint condition, For each individual, the corresponding formula parameters; The second step is the selection operation. The roulette wheel selection method is used to determine the probability of an individual being selected based on the proportion of an individual's fitness value in the total fitness of the population. Individuals with higher fitness values ​​have a greater probability of being selected, and individuals with better fitness performance are selected to enter the next generation. The third step is crossover operation; crossover operation is performed on the selected individuals according to the preset crossover probability, and the corresponding component parameters in the codes of the two individuals are swapped to generate new individuals. At the same time, it is ensured that the total component ratio of the new individuals meets the preparation requirements and there are no abnormal component ratios. The fourth step is mutation operation; for the population generated after crossover, the mutation operation is performed according to the preset mutation probability, randomly adjusting the proportion parameter of a certain component in the individual's code, generating new mutated individuals, expanding the search range, and avoiding the algorithm from getting stuck in a local optimum. The fifth step is non-dominated sorting and elite retention. Non-dominated sorting is performed on the new generation population to select the non-dominated individuals with the best fire resistance performance. The elite retention strategy is used to directly retain the best individuals from the previous generation into the new generation population to avoid the loss of the optimal solution. After each iteration, the optimal fitness value of the population is checked. If the change in the optimal fitness value of the population is less than the preset convergence threshold after 30 consecutive iterations, the algorithm is considered to have converged and the iteration process is terminated early. If the convergence condition is not met, the iteration steps continue to be executed in a loop until the preset maximum number of iterations is reached. After the algorithm terminates, the individual with the highest fitness value in the final population is extracted, and the corresponding formula is the optimal coating formula that satisfies all optimization objectives and constraints. At the same time, 3 to 5 suboptimal formulas are output as alternative optimal formulas in descending order of fitness value to provide redundant selection for subsequent preparation steps. For the output optimal formula and alternative preferred formula, conduct component rationality and preparation feasibility verification to confirm that the mass ratio of each component is within the range achievable by conventional preparation, and there are no component conflicts or unachievable processes. After verification, save the complete component mass ratio parameters, full-dimensional performance prediction data, and constraint condition satisfaction verification report of all preferred formulas to provide accurate proportioning basis and performance reference for the coating preparation in the subsequent step S5.

[0019] In this embodiment, the core function of step S5 is to build upon the optimized coating formulation determined in step S4. Using bamboo sap, a byproduct of bamboo carbonization, as the core base material, a standardized process of raw material pretreatment, graded dispersion, fine grinding, and performance regulation is employed to stabilize the bamboo sap-based fire-retardant coating. This ensures that the final coating has uniform components, stable performance, and perfectly matches the design ratio and performance expectations of the optimized formulation. This provides a suitable physical sample for subsequent performance verification tests and multi-dimensional digital twin model correction. The detailed steps are as follows: Step S5-1: Pre-treatment of bamboo sap water base material for purification: According to the design dosage of the optimized coating formula, the corresponding batch of bamboo sap water raw material was taken and pre-treated for purification. First, the bamboo sap water raw material was placed in a sealed container and allowed to settle at room temperature for no less than 12 hours to allow large solid particles in the raw material to settle to the bottom of the container. After settling, the upper clear liquid phase was extracted and filtered step by step using a multi-layer filter screen. The filtration process used 80-mesh, 150-mesh, and 200-mesh filter screens to complete three-stage filtration, thoroughly removing solid suspended matter and insoluble impurities from the bamboo sap water to obtain purified bamboo sap water. Basic performance tests were conducted on the purified bamboo sap water, including solids content, pH value and viscosity, to ensure that the performance parameters of the purified bamboo sap water were consistent with the performance parameters of the base material used to construct the coating performance parameter database in step S1. After the tests were completed, the purified bamboo sap water that met the requirements was sealed and temporarily stored to avoid volatilization and pollution, in order to prepare for subsequent material preparation. The formula for calculating solids content is: Where W represents the solids content of bamboo sap. The mass of the solid residue after drying to a constant weight. The quality of the bamboo sap sample before drying; Step S5-2: Primary low-speed dispersion and preparation of the first mixed slurry: The purified bamboo sap water, along with the formulated film-forming aid, dehydration catalyst, charring agent, foaming agent, and some other additives, are added to the dispersion vessel in the preset feeding order. The feeding order follows the principle of liquid phase first, then solid phase. First, all the purified bamboo sap water is added, followed by the liquid film-forming aid and other liquid additives. The dispersion vessel is started and stirred at a low speed of 300 rpm to 500 rpm for 5 to 10 minutes to ensure that the liquid phase components are fully mixed. Then, while continuously stirring, the dehydration catalyst, charring agent, and foaming agent powder components are added in batches to avoid clumping and dust caused by adding all the powder at once. After all components are added, adjust the stirring speed of the dispersion vessel to 800 rpm to 1200 rpm and maintain a normal temperature and pressure environment for continuous stirring for 20 to 40 minutes. During the stirring process, monitor the slurry state in real time to ensure no powder agglomeration or stratification, and ensure that all components are fully and uniformly mixed to obtain the first mixed slurry. After stirring, take multiple samples of the first mixed slurry and test the relative deviation of the solid content of samples at different locations to ensure that the mixing uniformity meets the requirements. The formula for calculating the mixing uniformity is: ,in, The coefficient of variation of the solid content in the slurry. The standard deviation of solid content detection values ​​from multiple sampling points. The arithmetic mean of the solid content detection values ​​from multiple sampling points; Step S5-3: Secondary high-speed dispersion and preparation of the second mixed slurry: Add the formulated amount of filler and remaining additives to the first mixed slurry. During the feeding process, keep the dispersion vessel stirring at a speed of 500 rpm to 800 rpm to ensure that the filler is evenly dispersed into the slurry system without local accumulation or clumping. After all fillers and additives have been added, increase the stirring speed of the dispersion vessel to 1500 rpm to 2500 rpm and keep the vessel in a closed environment for continuous high-speed dispersion for 30 to 60 minutes. During the high-speed dispersion process, the system temperature is controlled in real time by the jacketed cooling water system of the dispersion vessel to ensure that the slurry temperature does not exceed 40 degrees Celsius, thus avoiding the heat generated by high-speed shearing that could cause the volatilization of additives, denaturation of components, or a decrease in system stability. During the dispersion process, the slurry is sampled and observed every 10 minutes to confirm that the powder material is fully wetted and dispersed in the liquid phase and that there are no visible powder agglomerates. After the high-speed dispersion reaches the preset time, stirring is stopped to obtain the second mixed slurry. Step S5-4: Circulating fine grinding and slurry fineness control: The second mixed slurry is transferred to a horizontal sand mill for circulating fine grinding. Before grinding, the sand mill is debugged and prepared. Zirconia beads are used as the grinding media, with a particle size of 0.6mm to 1.2mm. The grinding media filling rate is controlled between 60% and 80%. The spindle speed of the sand mill is adjusted to 1000rpm to 1500rpm. The cooling water system of the sand mill jacket is turned on to control the temperature inside the grinding chamber to not exceed 45 degrees Celsius. The slurry is fed at a uniform rate, controlled between 50L / h and 100L / h. After entering the grinding chamber, the slurry is circulated and ground. After each full cycle, the fineness of the ground slurry is checked using a scraper fineness gauge. The grinding continues until the fineness of the slurry is less than or equal to 50μm. Grinding is stopped once the preset fineness requirement is met, and the ground slurry is obtained. After grinding, the ground slurry is fully filtered to remove any residual grinding media debris and large particles that are not fully dispersed, ensuring that the slurry system is uniform and stable. Step S5-5: Post-treatment control and preparation of finished fire-retardant coating: The ground slurry underwent final performance regulation and post-treatment to prepare the finished fire-retardant coating. First, the ground slurry was subjected to secondary precision filtration using a 200-mesh filter to thoroughly remove trace impurities and undispersed particles from the system. After filtration, the viscosity of the slurry was precisely regulated under low-speed stirring. By adding the rheology modifiers that match the formulation, the viscosity of the slurry was adjusted to the target range. Viscosity testing was performed using a Ford cup at room temperature (25 degrees Celsius), and the target viscosity range was set to 30 to 60 seconds. After viscosity adjustment, a full range of basic performance tests are conducted on the finished coating, including solid content, fineness, viscosity, storage stability, and workability. The storage stability test is completed by sealing the coating sample and placing it in a room temperature environment for 24 hours to confirm that the sample does not exhibit stratification, precipitation, clumping, or flocculation. After all test items meet the requirements, the finished fire-retardant coating is sealed and packaged, labeled with the corresponding formula number, preparation batch, preparation time, and performance parameters, completing the entire preparation process and providing a qualified physical sample for the subsequent performance verification test in step S6.

[0020] In this embodiment, the core function of step S6 is to take the finished fire-retardant coating prepared in step S5, obtain a full set of measured data on the actual performance of the coating through standardized fire resistance performance verification tests, and conduct quantitative comparison and deviation verification between the measured results and the corresponding formula performance prediction results output by the multi-dimensional digital twin model. For deviation items exceeding the preset threshold, key influencing parameters of the model are identified and calibrated. Then, the entire process of preparation and verification is optimized through closed-loop iterative simulation. Finally, the model prediction accuracy and the actual protective performance of the coating are both achieved, resulting in a final fire-retardant coating that is fully adapted to the target wood structure application scenario. This forms a closed-loop control system from digital simulation optimization to physical performance verification. The detailed steps are as follows: Step S6-1: Fire resistance performance verification test and verification data acquisition: Standard-sized samples were taken from the same batch of wood structure substrates used in the target wood structure application scenario. The sample specifications were completely matched with the national standard requirements for fire resistance testing. The fire-retardant coating prepared in step S5 was uniformly applied to the fire-exposed surface of the sample according to the standard construction process. The coating thickness was completely consistent with the initial coating thickness set in the simulation process in step S3. After coating, the sample was placed in a constant temperature and humidity environment for standard curing until the coating was completely cured. After curing, the surface condition of the coating was checked. Samples with cracks, peeling, pinholes or sagging defects were removed. Valid samples with qualified condition were selected for testing. The number of valid samples was not less than 3 groups. At the same time, blank samples of uncoated substrates from the same batch were prepared as performance control. The qualified coated sample is placed in a fire resistance test furnace, and a fire resistance test is carried out according to the preset fire scenario or heat exposure conditions set in step S3. The temperature rise curve and the convective heat transfer conditions of the furnace atmosphere during the test are completely consistent with the initial boundary conditions of the simulation, ensuring that the test results are directly comparable to the predicted results. During the test, a multi-channel data acquisition system is used to collect three core data in real time throughout the entire test cycle: the temperature of the sample's unexposed surface, the coating expansion thickness, and the temperature field distribution inside the substrate. The temperature of the unexposed surface is collected by thermocouples arranged at multiple points, and the thermocouples are arranged in accordance with the corresponding national standards, with a collection frequency of not less than 1 time per second. The coating expansion thickness is collected in real time by a high-precision laser displacement sensor, which synchronously records the dynamic changes in the coating thickness with the furnace temperature and the test time. The temperature field inside the substrate is collected by thermocouples embedded at different depths in the substrate, with the collection frequency consistent with the collection frequency of the unexposed surface temperature. The test continues until the preset termination condition is reached. The termination condition is completely consistent with the judgment criteria for the fire resistance limit in step S3, that is, the average temperature of the unexposed surface of the sample reaches the critical value or the sample shows penetrating failure. After the test is completed, the measured data of all valid parallel samples are preprocessed, and abnormal data groups with excessive dispersion coefficients are removed. The arithmetic mean of the remaining valid data is taken as the final measured result, and a verification dataset containing the measured fire resistance limit coating expansion ratio and the change of substrate pyrolysis depth over time is generated. The formula for calculating the coefficient of variation is: ,in, The coefficient of variation is the distance between the test data of parallel samples. The standard deviation of multiple sets of parallel sample test data. This is the arithmetic mean of multiple sets of parallel sample test data; Step S6-2: Time-series alignment and deviation quantification calculation of performance prediction and measured results: Using the experimental timeline of the validation dataset as a benchmark, the timeline of the performance prediction results of the corresponding candidate formulation in step S3 is precisely aligned with the measured timeline to ensure that the predicted values ​​and measured values ​​at the same time point correspond one-to-one and eliminate the misalignment error in the time dimension. After alignment, the measured full-cycle change curves of the refractory limit coating expansion ratio and the full-cycle change curves of the substrate pyrolysis depth in the validation dataset are compared item by item with the predicted curves of the refractory limit coating expansion behavior and the predicted curves of the substrate pyrolysis depth in the performance prediction results of the corresponding formulation. The absolute deviation, relative deviation and full-cycle fitting deviation of each core performance index are calculated respectively. The formula for calculating absolute deviation is: ,in, This represents the absolute deviation of a single indicator. This is the measured value of the indicator. This is the predicted value for the indicator; The formula for calculating relative deviation is: ,in, This represents the relative deviation of a single indicator. The fitting deviation of the full-cycle variation curve is quantified and calculated using the root mean square error. The calculation formula is as follows: ,in, The root mean square error of the full-cycle curve is... This represents the total number of data points collected throughout the entire cycle. For the first Measured values ​​at each time point For the first Predicted values ​​for each time point; After all deviation calculations are completed, a deviation analysis report for each core performance indicator is generated, which clarifies the deviation value and the direction of the deviation for each indicator, providing a quantitative basis for subsequent threshold verification and model correction. Step S6-3: Deviation threshold verification and process flow determination: A pre-set deviation threshold system is loaded, which includes the relative deviation threshold of each core performance indicator and the root mean square error threshold of the full-cycle curve. All thresholds are pre-determined based on the safety protection requirements of the target wood structure application scenario and the design accuracy requirements of the digital twin model. During the verification process, the calculated deviation of each core performance indicator is checked against the corresponding preset threshold to determine whether the deviation of all indicators is within the preset threshold range. If the deviations of all core performance indicators are within the preset threshold range, it is determined that the actual performance of the current fireproof coating is highly matched with the model prediction results and fully meets the design requirements of the target application scenario. The current fireproof coating is the final fireproof coating adapted to the target wood structure application scenario, and the whole process ends. If the deviation of any core performance indicator exceeds the preset threshold, it is determined that the prediction accuracy of the current model does not meet the design requirements, the actual performance of the current coating formulation deviates significantly from the expected performance, and it is necessary to proceed to the subsequent model parameter correction and closed-loop iteration stage. Step S6-4: Sensitivity analysis and inversion calibration correction of key model parameters: For performance indicators that exceed preset thresholds, a parameter sensitivity analysis of a multi-dimensional digital twin model is conducted to identify key model parameters whose influence on the performance indicator exceeds a set contribution threshold. The sensitivity analysis is carried out using the controlled variable method, keeping all other parameters in the model unchanged except for the parameter to be analyzed. Within a reasonable range of parameter values, the values ​​of individual parameters are adjusted one by one, and the change in the predicted result of the target performance indicator corresponding to the unit change of the parameter is calculated to quantify the sensitivity coefficient of each parameter. The formula for calculating the sensitivity coefficient is: ,in, Let be the sensitivity coefficient of the parameter to be analyzed. The change in the predicted result of the target performance index. The initial value for the predicted target performance index. The adjustment amount for the parameter to be analyzed. These are the initial values ​​for the parameters to be analyzed; A contribution threshold for the sensitivity coefficient is set, and parameters with sensitivity coefficients greater than the contribution threshold are identified as key model parameters that significantly affect the deviation term. Based on the measured data of the validation dataset, the Levenberg-Marquardt parameter inversion algorithm is used to calibrate and correct the identified key model parameters. The parameter inversion process takes minimizing the total deviation between the model simulation results and the measured data as the optimization objective, and iteratively solves for the optimal calibration value of the key model parameters, so that the deviation between the simulation results output by the corrected multi-dimensional digital twin model and the corresponding measured data of the validation dataset is reduced to within the preset threshold range. After the parameters are corrected, the accuracy of the calibrated model is verified. The experimental boundary conditions corresponding to the verification dataset are input into the corrected model, the simulation calculation is carried out again and the performance prediction results are output. The deviation between the new prediction results and the measured data is checked, and it is confirmed that the deviation of all core performance indicators meets the preset threshold requirements. The parameter calibration and correction of the multi-dimensional digital twin model is completed. Step S6-5: Model Update and Full-Process Closed-Loop Iterative Optimization: The multi-dimensional digital twin model, after parameter calibration and correction, is used as the updated standard model. The entire process is then repeated in step S3. First, all candidate bamboo sap water-based fire-retardant coating formulations are input into the updated multi-dimensional digital twin model, and multi-physics coupled transient simulation is performed again to output a new set of performance prediction results for each candidate formulation corresponding to the updated model. Then, in step S4, multi-objective constrained formulation optimization is performed based on the new performance prediction result set to select a new preferred coating formulation that fits the updated model. Next, in step S5, the standardized preparation of bamboo sap water-based fire-retardant coating is completed according to the new preferred coating formulation. Finally, in step S6, the fire resistance performance of the newly prepared coating is verified and deviations are checked. Repeat the above closed-loop iterative process, completing all stages of model update, simulation calculation, formula optimization, coating preparation, and performance verification in each iteration, until the deviations of all core performance indicators in step S6-3 are stably within the preset threshold range. After the iteration terminates, the final fire-retardant coating is the final fire-retardant coating fully adapted to the target wood structure application scenario. Simultaneously, the full-process simulation data and performance verification test data of the multi-dimensional digital twin model after the final determined coating complete formula calibration are saved, forming a complete and traceable technical archive, providing complete technical support for the subsequent batch preparation and engineering application of coatings in the same scenario. Although embodiments of the present invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for preparing fire-retardant coatings based on multi-dimensional digital twins, characterized in that: Includes the following steps: S1. Collect time-series data of environmental parameters of the target wood structure application scenario, physical performance parameters of wood structure substrate, and performance parameter database of fireproof coatings based on bamboo sap, a by-product of bamboo carbonization. S2. Construct a digital twin model of the environmental field based on time-series data of environmental parameters; construct a digital twin model of the thermal response of the wood structure based on the physical performance parameters of the wood structure substrate and according to the heat conduction equation and the pyrolysis reaction kinetic equation; and construct a digital twin model of the coating performance based on the performance parameter database using a data-driven method. The environmental boundary conditions output from the digital twin sub-model of the environmental field are loaded into the digital twin sub-model of the thermodynamic response of the wood structure, and the dynamic thermophysical property parameters of the coating output from the digital twin sub-model of the coating performance are used as the surface boundary conditions of the digital twin sub-model of the thermodynamic response of the wood structure. This achieves multi-physics coupling of environment, substrate and coating, and establishes a multi-dimensional digital twin model. S3. Input multiple candidate bamboo sap water-based fireproof coating formulations into a multi-dimensional digital twin model and perform multi-physics field coupled transient simulation to simulate the temperature field evolution, carbonization layer formation and growth process of the wood structure under preset fire scenarios or heat exposure conditions. Output the performance prediction results of the fire resistance limit, coating expansion behavior and substrate pyrolysis depth change with time for each candidate formulation. S4. Based on the preset optimization objectives, an optimization algorithm is used to find the best coating formulation among multiple candidate formulations and select the preferred coating formulation that meets the optimization objectives. The optimization objectives include maximizing the fire resistance limit, ensuring that the coating expansion ratio is within the set range, and ensuring that the coating adhesion meets the standard. S5. According to the preferred coating formula, bamboo sap is mixed with film-forming aid, dehydration catalyst, charring agent, foaming agent, filler and additives in proportion, and then dispersed and ground to obtain fireproof coating. S6. The performance of the prepared fireproof coating is verified. If the deviation between the verification result and the performance prediction result of the corresponding formula exceeds the preset threshold, the parameters of the multi-dimensional digital twin model are corrected according to the verification result. Steps S3 to S5 are repeated based on the corrected model until the deviation between the verification result and the performance prediction result is within the preset threshold, and the final fireproof coating adapted to the target wood structure application scenario is obtained.

2. The method for preparing fire-retardant coatings based on multi-dimensional digital twins according to claim 1, characterized in that: The environmental parameters time series data include at least temperature, humidity and solar radiation intensity; the physical performance parameters of the wood structure substrate include at least density, specific heat capacity, thermal conductivity and pyrolysis kinetic parameters; the performance parameter database contains the thermal conductivity, specific heat capacity, expansion ratio, char formation rate and initial expansion temperature of fireproof coatings under different formulations.

3. The method for preparing fire-retardant coatings based on multi-dimensional digital twins according to claim 2, characterized in that: The database collects time-series data of environmental parameters for the target wood structure application scenario, physical performance parameters of the wood structure substrate, and performance parameters of fire-retardant coatings based on bamboo sap, a byproduct of bamboo carbonization. Specifically, this includes: S1-1. Deploy multiple environmental sensor nodes at typical locations in the target timber structure application scenario to collect real-time time-series data of temperature, humidity and solar radiation intensity, and perform outlier removal and missing value imputation on the collected raw data to obtain a standardized environmental parameter time-series dataset. S1-2. Drill core samples from the wood structure substrate of the target wood structure application scenario, determine the density, specific heat capacity and thermal conductivity of the core samples using a thermal constant analyzer, and determine the pyrolysis kinetic parameters of the core samples using a thermogravimetric analyzer to establish a set of physical performance parameters for the wood structure substrate. S1-3. Using bamboo sap as a base material, multiple candidate formulation coating samples were prepared by mixing with film-forming aids, dehydration catalysts, charring agents, foaming agents, fillers and additives in different proportions. Thermogravimetric analysis, thermal conductivity testing and refractory chamber tests were performed on each sample to determine the thermal conductivity, specific heat capacity, expansion ratio, charring rate and initial expansion temperature of each formulation, and a coating performance parameter database was constructed.

4. The method for preparing fire-retardant coatings based on multi-dimensional digital twins according to claim 1, characterized in that: Step S2 specifically includes: S2-1. Using a standardized environmental parameter time series dataset as input, a time series modeling method is used to construct an environmental field digital twin model to describe the spatiotemporal distribution and dynamic changes of temperature, humidity and radiation fields in the target timber structure application scenario. S2-2. Using the set of physical performance parameters of the wood structure substrate as input, a heat conduction equation describing the evolution of the internal temperature field of the wood structure is established based on Fourier's heat conduction law, and a pyrolysis reaction kinetic equation describing the pyrolysis reaction rate of the wood structure material is established based on Arrhenius's law. The heat conduction equation and the pyrolysis reaction kinetic equation are coupled to construct a digital twin model of the wood structure thermodynamic response that can simulate the dynamic evolution of the internal temperature distribution and pyrolysis degree of the wood structure under heating conditions. S2-3. Using the coating performance parameter database as training samples, deep learning algorithms are used to train the model and establish a digital twin model of coating performance that can dynamically map the changes in thermal conductivity, specific heat capacity, expansion ratio and char rate of the corresponding coating during the entire heating process based on the coating formulation parameters. S2-4. The environmental boundary conditions, including air temperature, convective heat transfer coefficient and radiative heat flux, calculated by the digital twin sub-model of the environmental field, are applied as external loads to the uncoated surface of the digital twin sub-model of the thermal response of the wood structure. S2-5. The dynamic thermophysical property parameters of the coating, which change in real time with the coating temperature and degree of pyrolysis, output by the digital twin sub-model of coating performance, are applied as surface boundary conditions to the coated surface of the digital twin sub-model of wood structure thermal response. This realizes the multi-physics coupling between the environmental field, coating performance and wood structure thermal response, thereby establishing a multi-dimensional digital twin model for simulating the comprehensive thermal response behavior of wood structures coated with fire-retardant coatings under real environmental and fire conditions.

5. The method for preparing fire-retardant coatings based on multi-dimensional digital twins according to claim 1, characterized in that: Step S3 specifically includes: S3-1. Input multiple candidate bamboo sap water-based fireproof coating formulations into a multi-dimensional digital twin model, construct a corresponding formulation parameter set for each candidate formulation, and use it as the input variable of the coating performance digital twin sub-model. S3-2. Set a unified preset fire scenario or thermal exposure conditions as the initial boundary conditions for simulation, and load them into the digital twin sub-model of the environmental field to generate the corresponding time-varying thermal environment parameters. S3-3, The multi-dimensional digital twin model is run, based on the time-varying thermal environment parameters and recipe parameter set, to perform multi-physics field coupled transient simulation, and calculate and record the temperature field distribution data of the wood structure at different simulation times in real time; S3-4. Based on the temperature field distribution data, the isothermal surface movement trajectory of the wood structure surface and interior reaching the pyrolysis temperature threshold is identified and tracked through the digital twin model of the wood structure thermal response, simulating the formation and growth process of the carbonized layer. S3-5. Extract the key performance indicators corresponding to each candidate formulation from the simulation results, including the time required for the temperature on the back of the wood structure to reach the critical value as the fire resistance limit, the dynamic curve of the coating thickness changing with temperature as the coating expansion behavior, and the change of the position of the carbonized layer front relative to the original surface over time as the pyrolysis depth of the substrate. S3-6. Perform time-series alignment and feature quantification on the extracted key performance indicators to generate a performance prediction result set that includes the changes in fire resistance limit, coating expansion behavior and substrate pyrolysis depth over time, and store it in association with the corresponding candidate formulations.

6. The method for preparing fire-retardant coatings based on multi-dimensional digital twins according to claim 1, characterized in that: Based on the preset optimization objectives, an optimization algorithm is used to find the best coating formulation among multiple candidate formulations and select the preferred formulation that meets the optimization objectives. The optimization objectives include maximizing the fire resistance limit, ensuring the coating expansion ratio is within a set range, and achieving the required coating adhesion, specifically including: S4-1. Extract the predicted value of the fire resistance limit, the predicted value of the maximum expansion ratio of the coating during the expansion process, and the measured value of the coating adhesion corresponding to the formula from the performance prediction results of each candidate formula output in step S3, and construct a comprehensive dataset of candidate formulas containing performance prediction data and measured data. S4-2. Based on the comprehensive dataset of candidate formulations, construct a multi-objective constrained optimization space according to the preset optimization objectives. Among them, maximizing the fire resistance limit is taken as the main optimization index, and the coating expansion ratio being within the set range and the coating adhesion meeting the standard are taken as the constraint conditions. S4-3. The constraint that the coating expansion ratio is within a set range is transformed into a numerical range screening of the coating expansion behavior curve in the performance prediction results. Candidate formulations with the predicted maximum expansion ratio of the coating being lower than the lower threshold or higher than the upper threshold are eliminated to form the first set of screening formulations that meet the expansion ratio constraint. S4-4. The constraint condition for achieving the coating adhesion standard is transformed into a level screening of the measured adhesion value. Candidate formulations whose measured adhesion values ​​do not meet the preset level standard are removed from the first screening formulation set to form a second screening formulation set that simultaneously satisfies the expansion ratio constraint and the adhesion constraint. S4-5. Optimization algorithms are used to calculate the candidate formulations in the second screening formulation set. Iterative optimization is carried out with the goal of maximizing the predicted fire resistance limit. One or more candidate formulations with the best predicted fire resistance limit are selected from the second screening formulation set as the preferred coating formulations that meet the optimization objectives.

7. The method for preparing fire-retardant coatings based on multi-dimensional digital twins according to claim 1, characterized in that: According to the optimized coating formula, bamboo sap is mixed with film-forming aids, dehydration catalysts, charring agents, foaming agents, fillers, and additives in a specified ratio. After dispersion and grinding, a fire-retardant coating is obtained, specifically comprising: S5-1. The required amount of bamboo sap water in the preferred coating formula is filtered and purified to remove solid impurities and obtain purified bamboo sap water. S5-2. The purified bamboo sap water, film-forming aid, dehydration catalyst, charring agent, foaming agent and some of the additives are added to the dispersion vessel in sequence and stirred at a speed of 800-1200 rpm for 20-40 minutes to make the components fully mixed and uniform, so as to obtain the first mixed slurry. S5-3. Add the formulated amount of filler and remaining additives to the first mixed slurry, increase the speed of the dispersion kettle to 1500-2500 rpm, and continue high-speed dispersion for 30-60 minutes to fully wet and disperse the powder material in the liquid phase to obtain the second mixed slurry. S5-4. Transfer the second mixed slurry to a horizontal sand mill for grinding, control the grinding media filling rate and grinding time, until the fineness of the ground slurry is ≤50μm, and obtain the ground slurry; S5-5. Filter the grinding slurry to remove the grinding media and undispersed particles, and adjust the viscosity to the target range to obtain the fireproof coating.

8. The method for preparing fire-retardant coatings based on multi-dimensional digital twins according to claim 1, characterized in that: The prepared fire-retardant coating undergoes performance verification. If the deviation between the verification result and the performance prediction result of the corresponding formulation exceeds a preset threshold, the parameters of the multi-dimensional digital twin model are corrected based on the verification result. Steps S3 to S5 are repeated based on the corrected model until the deviation between the verification result and the performance prediction result is within the preset threshold, thus obtaining the final fire-retardant coating suitable for the target wood structure application scenario. Specifically, this includes: S6-1. Apply the fire-retardant coating prepared according to the preferred coating formula to the target wood structure substrate sample according to the standard construction process. After the coating is cured, place the coated sample in the fire resistance test furnace and conduct a fire resistance test according to the preset fire scenario or heat exposure conditions set in step S3. Collect the temperature of the unexposed surface of the sample, the coating expansion thickness and the temperature field distribution inside the substrate in real time during the test to obtain the measured fire resistance limit, coating expansion ratio and substrate pyrolysis depth change over time verification dataset. S6-2. Compare the measured fire resistance limit, coating expansion ratio and substrate pyrolysis depth curves in the verification dataset with the corresponding predicted values ​​of fire resistance limit, coating expansion behavior and substrate pyrolysis depth in the performance prediction results of the corresponding candidate formulation in step S3, and calculate the absolute or relative deviation of each performance index. S6-3. Determine whether the deviations of each performance indicator are all within the preset threshold range: If they are all within the preset threshold, determine that the current fireproof coating is the final fireproof coating suitable for the target wood structure application scenario and end the process; if the deviation of any performance indicator exceeds the preset threshold, proceed to step S6-4. S6-4. Based on the performance indicators that exceed the preset threshold, the key model parameters in the multi-dimensional digital twin model that have an impact on the corresponding performance indicators exceeding the set contribution level are identified through sensitivity analysis. Based on the measured data in the verification dataset, the identified key model parameters are calibrated and corrected using a parameter inversion algorithm, so that the deviation between the simulation results output by the corrected multi-dimensional digital twin model and the corresponding measured data in the verification dataset is reduced to within the preset threshold range. S6-5. Using the multi-dimensional digital twin model with corrected parameters as the updated model, return to step S3 and re-execute the multi-physics coupled transient simulation, formula optimization, and coating preparation and verification steps of the candidate formula until the deviations of each performance index in step S6-3 are within the preset threshold, and finally obtain the fireproof coating that is suitable for the target wood structure application scenario.

Images