Material proportioning intelligent optimization method and device fusing domain knowledge and machine learning

Abstract

The application is applicable to the field of building materials, and provides a material proportioning intelligent optimization method and device integrating field knowledge and machine learning, multiple sets of training data sets of fiber cement-based materials are obtained, each set of training data set at least includes training fiber data; a target first regression coefficient of the fiber cement-based material is determined based on the multiple sets of training data sets and a compressive strength formula of the cement-based material; a first target prediction model of the fiber cement-based material is constructed according to the multiple sets of training data sets and the target first regression coefficient; a second target prediction model of the fiber cement-based material is constructed, and the mix proportion of the mixed material in the fiber cement-based material is calculated. The method integrates the fiber data into the compressive strength formula to determine two regression coefficients suitable for the fiber cement-based material, and then constructs a prediction model for predicting the compressive strength of the fiber cement-based material by using the determined two regression coefficients, so that the influence of fiber mixing is considered when calculating the material mix proportion, and the design precision of the mix proportion is improved.

Description

Technical Field

[0001] This application belongs to the field of building materials technology, and in particular relates to a method and device for intelligent optimization of material proportions that integrates domain knowledge and machine learning. Background Technology

[0002] Fiber cement-based materials, as a new type of building material, are a general term for composite materials composed of fiber and cement-based materials. Fiber cement-based materials can be considered as materials composed of cement, mortar, etc. as base materials and fiber as reinforcing material. They can significantly improve the ductility, crack resistance and toughness of traditional cement-based materials and have important application value in engineering scenarios such as earthquake resistance and impact resistance.

[0003] However, the performance of fiber cementitious materials is influenced by a combination of factors, including the composition of the base material (such as cement, fly ash, mineral powder, etc.), fiber properties (such as type, length, volume fraction, etc.), and mix proportion parameters (such as water-cement ratio, sand-cement ratio, etc.), resulting in a complex nonlinear relationship.

[0004] In traditional methods for determining the mix proportions of fiber cementitious materials, the regression coefficients are mostly empirical or fixed values. For example, directly using the regression coefficients determined by experimental data of traditional cementitious materials to determine the mix proportions of fiber cementitious materials does not take into account the influence of fiber incorporation on the mix proportions, resulting in insufficient accuracy in mix proportion design and making it difficult to achieve precise control over the performance of fiber cementitious materials. Summary of the Invention

[0005] This application provides a method and apparatus for intelligent optimization of material proportions that integrates domain knowledge and machine learning. By considering the influence of fiber incorporation on the material proportions, the accuracy of the proportion design is improved, and precise control of the performance of fiber cementitious materials is achieved.

[0006] In a first aspect, embodiments of this application provide a method for intelligent optimization of material proportions that integrates domain knowledge and machine learning, including:

[0007] Obtain multiple training datasets for fiber cement-based materials, wherein each training dataset includes at least: training cement ratio, training mortar-binder ratio, and training fiber data for fiber cement-based materials;

[0008] Based on multiple training datasets and the compressive strength formula of cement-based materials, the target first regression coefficient of fiber cement-based materials is determined. The compressive strength formula is determined based on at least the first and second regression coefficients of cement-based materials. The target first regression coefficient is used to calculate the actual compressive strength of fiber cement-based materials.

[0009] According to the multiple sets of training data sets and the target first regression coefficient, a first target prediction model of the fiber cement-based material is constructed, the first target prediction model is used to predict an actual second regression coefficient of the fiber cement-based material, and the actual second regression coefficient is used to calculate an actual compressive strength of the fiber cement-based material.

[0010] According to the multiple sets of training data sets and the first target prediction model, a second target prediction model of the fiber cement-based material is constructed, and the second target prediction model is used to predict an actual compressive strength of the fiber cement-based material.

[0011] Based on a target function and a constraint condition of the fiber cement-based material, a mix proportion of a mixed material in the fiber cement-based material is calculated, and at least part of the target function is determined based on the actual compressive strength output by the second target prediction model and the target compressive strength.

[0012] In a second aspect, an embodiment of the present application provides a material mix intelligent optimization device fusing field knowledge and machine learning, comprising:

[0013] The acquisition module is configured to acquire multiple sets of training data sets of the fiber cement-based material, wherein each set of training data set at least includes training cement ratio, training sand glue ratio and training fiber data of the fiber cement-based material.

[0014] The regression coefficient determination module is configured to determine a target first regression coefficient of the fiber cement-based material based on the multiple sets of training data sets and a compressive strength formula of the cement-based material, and the compressive strength formula is determined based on at least a first regression coefficient and a second regression coefficient of the cement-based material, and the target first regression coefficient is used to calculate an actual compressive strength of the fiber cement-based material.

[0015] The first model construction module is configured to construct a first target prediction model of the fiber cement-based material according to the multiple sets of training data sets and the target first regression coefficient, and the first target prediction model is used to predict an actual second regression coefficient of the fiber cement-based material, and the actual second regression coefficient is used to calculate an actual compressive strength of the fiber cement-based material.

[0016] The second model construction module is configured to construct a second target prediction model of the fiber cement-based material according to the multiple sets of training data sets and the first target prediction model, and the second target prediction model is used to predict an actual compressive strength of the fiber cement-based material.

[0017] The calculation module is configured to calculate a mix proportion of a mixed material in the fiber cement-based material based on a target function and a constraint condition of the fiber cement-based material, and at least part of the target function is determined based on an actual compressive strength output by the second target prediction model and a target compressive strength.

[0018] In a third aspect, an embodiment of the present application provides a terminal device, comprising a memory, a processor, and a computer program stored in the memory and capable of running on the processor, and the processor implements the method of any one of the first aspect when executing the computer program.

[0019] In a fourth aspect, an embodiment of the present application provides a computer readable storage medium, which stores a computer program, and the computer program is executed by a processor to implement the method of any one of the first aspect.

[0020] In a fifth aspect, an embodiment of the present application provides a computer program product, which, when running on a terminal device, causes the terminal device to execute the method of any one of the first aspect.

[0021] The embodiment of the present application provides a material proportioning intelligent optimization method and device fusing field knowledge and machine learning. The method comprises the following steps: obtaining a plurality of training data sets of a fiber cement-based material, wherein each training data set at least comprises training cement ratio, training sand glue ratio and training fiber data of the fiber cement-based material; determining a target first regression coefficient of the fiber cement-based material based on the plurality of training data sets and a compressive strength formula of the cement-based material, the compressive strength formula being determined based on at least a first regression coefficient and a second regression coefficient of the cement-based material, and the target first regression coefficient being used to calculate an actual compressive strength of the fiber cement-based material; constructing a first target prediction model of the fiber cement-based material according to the plurality of training data sets and the target first regression coefficient, the first target prediction model being used to predict an actual second regression coefficient of the fiber cement-based material, and the actual second regression coefficient being used to calculate the actual compressive strength of the fiber cement-based material; constructing a second target prediction model of the fiber cement-based material according to the plurality of training data sets and the first target prediction model, the second target prediction model being used to predict the actual compressive strength of the fiber cement-based material; and calculating a mix proportion of a mixed material in the fiber cement-based material based on a target function and a constraint condition of the fiber cement-based material, wherein at least part of the target function is determined based on the actual compressive strength output by the second target prediction model and a target compressive strength. By using the above technical solution, two regression coefficients suitable for the fiber cement-based material are determined by fusing the fiber data into the compressive strength formula, and then the prediction model for predicting the compressive strength of the fiber cement-based material is constructed by using the determined two regression coefficients, so that the influence of fiber incorporation is considered when calculating the material mix proportion, the design precision of the mix proportion is improved, and the performance of the fiber cement-based material is accurately controlled. BRIEF DESCRIPTION OF DRAWINGS

[0022] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed to be used in the embodiments or prior art description will be briefly introduced as follows. Obviously, the drawings in the following description only some embodiments of the present application, and for those skilled in the art, other drawings can also be obtained without creative labor on the basis of these drawings.

[0023] Figure 1 is a process schematic diagram of a material proportioning intelligent optimization method fusing domain knowledge and machine learning provided by an embodiment of the present application;

[0024] Figure 2 is a process schematic diagram of a material proportioning intelligent optimization method fusing domain knowledge and machine learning provided by another embodiment of the present application;

[0025] Figure 3 is a comparison schematic diagram of a true value and a predicted value of a second regression coefficient provided by an embodiment of the present application;

[0026] Figure 4 is a comparison schematic diagram of a true value and a predicted value of a second regression coefficient provided by an embodiment of the present application;

[0027] Figure 5 is a structure block diagram of a material proportioning intelligent optimization device fusing domain knowledge and machine learning provided by an embodiment of the present application;

[0028] Figure 6 is a structure schematic diagram of a terminal device provided by an embodiment of the present application. DETAILED DESCRIPTION

[0029] In the following description, specific details are set forth in order to provide a thorough understanding of embodiments of the present application. However, persons of ordinary skill in the art will readily recognize that embodiments of the present application can be practiced without these specific details, in other instances, well-known structures, devices, circuits, and methods have not been described in detail in order to avoid obscuring the present application.

[0030] It should be understood that, when used in the specification and the appended claims of the present application, the term "comprising" indicates the presence of the described features, integers, steps, operations, elements, and / or components, but does not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or groups thereof.

[0031] It should also be appreciated that the term "and / or" as used herein is used to associate together alternatively occurring events or circumstances that have a functional relationship described by the context of the present application and the associated claims. By contrast, the term "and" as used herein is used to conjoin together events or circumstances that are all simultaneously occurring and that have a functional relationship described by the context of the present application and the associated claims. Similarly, the term "or" as used herein is used to associate together alternatives such that the term "one or the other or both" is intended. By contrast, the term "either" as used herein is used to conjoin together events or circumstances that are all simultaneously occurring and that have a functional relationship described by the context of the present application and the associated claims. The term "one of" as used herein is used to conjoin together events or circumstances that are all simultaneously occurring and that have a functional relationship described by the context of the present application and the associated claims. The term "one or the other but not both" is intended to mean that only one of the associated events or circumstances can occur at any given time. The term "and / or" as used herein is used to associate together alternatively occurring events or circumstances that have a functional relationship described by the context of the present application and the associated claims, and includes all possible combinations of the associated events or circumstances.

[0032] As used in the description of the application and the appended claims, the term "if' can be interpreted to mean "when" or "upon" or "in response to determining" or "in response to detecting" depending on the context. Similarly, the phrase "if it is determined" or "if [a described condition or event] is detected" can be interpreted to mean "once it is determined" or "in response to determining" or "once [the described condition or event] is detected" or "in response to detecting [the described condition or event]," depending on the context.

[0033] In addition, in the description of the application and the appended claims, the terms "first", "second", "third", etc. are used merely to identify corresponding components without imposing or implying any relative importance.

[0034] Reference throughout this specification to "one embodiment" or "an embodiment" or "a specific embodiment" or "some embodiments" means that a particular feature, structure, or characteristic described in connection with the embodiment is included in at least one embodiment of the application. Thus, appearances of the phrases "in one embodiment" or "in some embodiments" or "in other embodiments" or "in additional embodiments" or the like in various places throughout this specification are not necessarily all referring to the same embodiment, unless otherwise indicated. Furthermore, the terms "comprises", "comprising", "includes", "including", "has", "having" and the like are intended to be open-ended terms that do not exclude additional, unrecited elements or methods. The term "coupled" is intended to mean a direct or indirect connection between two or more elements, and can include any combination of a direct connection, a wireless connection, and / or an indirect connection through one or more additional elements and / or interfaces.

[0035] It can be considered that the traditional method of determining the material mix ratio of fiber cement-based materials has the following limitations:

[0036] 1. Limitation of application of domain knowledge: In the traditional compressive strength formula of cement-based materials, the regression coefficient does not fully consider the influence of fiber incorporation on the microstructure and macroscopic mechanical properties of the material, making it difficult to accurately predict the performance of fiber cement-based materials with the theoretical model. In addition, the core regression coefficient in the existing formula is usually a fixed value, which cannot be dynamically adjusted according to the type of fiber, the dosage, and the mix ratio of key parameters, resulting in a significant deviation between the theoretical calculation results and the actual performance. Cement-based materials can be understood as a type of composite material formed by mixing cement as the main cementitious material with water, aggregates (sand, stone, etc.) and admixtures. Cement-based materials are one of the most widely used materials in modern construction engineering, including concrete, mortar, cement paste, etc.

[0037] 2. Defects of pure data-driven models: In recent years, machine learning algorithms such as neural networks and random forests have been used for performance prediction of fiber cement-based materials, but these machine learning algorithms rely on a large amount of labeled data and are prone to "overfitting" or "lack of physical meaning". For example, pure data-driven models may produce prediction results that violate basic material mechanics laws (such as strength increasing without limit as water-binder ratio increases), and pure data-driven models have poor interpretability and are difficult to guide mix proportion optimization.

[0038] 3. Complexity of multi-objective optimization: Fiber cement-based material design needs to meet the comprehensive needs of strength (28-day compressive strength), economy (material cost), and low carbon (carbon footprint). Traditional methods for determining the material mix proportion of fiber cement-based materials mainly include trial-and-error methods or single-objective optimization methods, which are difficult to balance the trade-off relationship between multiple objectives. For example, simply pursuing high strength will lead to an increase in cement usage, which not only increases costs but also increases carbon emissions, which does not meet the development trend of green buildings.

[0039] 4. Lack of dynamic correction of parameters: In traditional methods for determining the material mix proportion of fiber cement-based materials, key parameters affecting compressive strength (such as core regression coefficients) are mostly empirical values or fixed values, without considering the dynamic influence of fiber properties (such as diameter, aspect ratio, and volume fraction) and component proportions of the base material on compressive strength, resulting in insufficient design precision of the material mix proportion and difficulty in achieving precise control of material performance.

[0040] Based on this, the embodiments of the present application provide a material mix proportion intelligent optimization method that combines domain knowledge and machine learning, which can predict by constraining the model with physical laws, dynamically correct key parameters, and optimize the trade-off relationship between multiple objectives, achieving efficient, precise, and green design of fiber cement-based materials.

[0041] Figure 1 is a flowchart of a material mix proportion intelligent optimization method that combines domain knowledge and machine learning provided by an embodiment of the present application, which is an example and not a limitation. This method can be applied to terminal devices such as Figure 1 As shown in the figure, the method includes:

[0042] S101, obtaining multiple sets of training data sets of fiber cement-based materials.

[0043] Each set of training data set at least includes training cement ratio, training sand-binder ratio, and training fiber data of the fiber cement-based material. Other types of training data can also be included based on the above training data to enrich the content of the training data set. Each set of training data set can also include, for example, training water-binder ratio, training cement strength, and training compressive strength.

[0044] S102, determine a target first regression coefficient of the fiber cement-based material based on the multiple sets of training data sets and the compressive strength formula of the cement-based material.

[0045] The compressive strength formula is determined based on at least the first regression coefficient and the second regression coefficient of the cement-based material, and the target first regression coefficient is used to calculate the actual compressive strength of the fiber cement-based material.

[0046] It should be noted that the first regression coefficient and the second regression coefficient in the existing compressive strength formula of the cement-based material are obtained according to the experimental data of the cement-based material, so the first regression coefficient and the second regression coefficient are mainly suitable for calculating the compressive strength of the cement-based material, and are not suitable for calculating the compressive strength of the fiber cement-based material. Therefore, this step can first statistically analyze and parameter fit the multiple sets of training data sets of the fiber cement-based material to fit a set of regression coefficients suitable for the fiber cement-based material, and use the fitted regression coefficients to correct the compressive strength formula of the cement-based material, so that the corrected compressive strength formula can adapt to the characteristics of the fiber cement-based material, and provide a data basis for subsequent improvement of the prediction effect of the model.

[0047] The fitted regression coefficient may, for example, include a target first regression coefficient, i.e., a first regression coefficient suitable for the fiber cement-based material. The process of specifically determining the target first regression coefficient is not limited here, such as can be assisted by a pre-configured neural network model, by directly inputting the multiple sets of training data sets and the compressive strength formula of the cement-based material into the neural network model, to make the neural network model output the corresponding target first regression coefficient, or can also set initial values for the first regression coefficient and the second regression coefficient, and use an iterative optimization algorithm to continuously correct and iterate the initial values of the first regression coefficient and the second regression coefficient to obtain the target first regression coefficient.

[0048] As a feasible implementation manner, the target first regression coefficient of the fiber cement-based material is determined based on the multiple sets of training data sets and the compressive strength formula of the cement-based material, comprising:

[0049] The first initial value is determined as the current first regression coefficient of the fiber cement-based material, and the second initial value is determined as the current second regression coefficient of the fiber cement-based material;

[0050] For each set of training data sets, the training water-binder ratio, the training cement strength, the current first regression coefficient and the current second regression coefficient of each set of training data sets are input into the compressive strength formula to obtain the theoretical compressive strength of each set of training data sets;

[0051] Based on the theoretical compressive strength and the training compressive strength of each set of training data sets, the current residual sum of squares of the multiple sets of training data sets is calculated;

[0052] According to the current residual sum of squares, the current first regression coefficient and the current second regression coefficient are modified, and the step of inputting the training water-binder ratio, the training cement strength, the current first regression coefficient and the current second regression coefficient of each training data set into the compressive strength formula to obtain the theoretical compressive strength of each training data set is returned until the current residual sum of squares reaches convergence or the number of modifications of the current first regression coefficient reaches the maximum iteration number.

[0053] The current first regression coefficient corresponding to the current residual sum of squares reaching convergence or the number of modifications of the current first regression coefficient reaching the maximum iteration number is determined as the target first regression coefficient of the fiber cement-based material.

[0054] The first initial value and the second initial value can be understood as the iteration starting point of the iterative optimization algorithm, which helps the algorithm converge to the global optimal solution faster and avoids fitting failure or result deviation caused by unreasonable initial values. For example, if the initial value is too different from the true value, the iteration will not converge (error) or meaningless parameters will be obtained. The specific content of the first initial value and the second initial value can be determined according to the experience value.

[0055] Optionally, the iterative optimization algorithm can be the Levenberg-Marquardt (LM) algorithm, and the first initial value and the second initial value can be 2 and 1 respectively. The current first regression coefficient and the current second regression coefficient are the first regression coefficient and the second regression coefficient corresponding to each current iteration period.

[0056] Further, the compressive strength formula can be wherein, represents the compressive strength, represents the cement strength, W / B represents the water-binder ratio, respectively represents the first regression coefficient and the second regression coefficient. It can be seen that the compressive strength formula is a nonlinear model because the parameters exist a multiplication relationship, which means that the relationship between the parameters and the dependent variable cannot be expressed by a simple linear combination.

[0057] In the specific embodiment, the above compressive strength formula is first determined as the model function fcu_model, and it can be known that is an unknown parameter to be solved and is also the target to be fitted.

[0058] Then the training water-binder ratio and training cement strength of multiple training data sets can be combined into a two-dimensional array with shape (2, n) by np.vstack, which meets the requirements of the curve_fit function for multiple independent variable inputs, and n is the number of training data sets. Among them, the curve_fit function can be used for nonlinear least squares fitting.

[0059] Then the curve_fit function can find the optimal parameters by minimizing the "residual sum of squares of real value and model predicted value". The specific steps can include: in the first iteration round, the two-dimensional array, the first initial value and the second initial value initial value [2, 1] can be substituted into the model fcu_model, and the predicted value (i.e. theoretical compressive strength) corresponding to each training data set can be calculated; then the residual between the predicted value and the real value (i.e. training compressive strength) is calculated, and the current residual sum of squares of multiple training data sets is solved; then the optimization algorithm (such as least squares method) inside the curve_fit function is called, and the value of is corrected according to the current residual sum of squares; then in the subsequent iteration rounds, the two-dimensional array, the corrected can be substituted into the model fcu_model, and the predicted value (i.e. theoretical compressive strength) corresponding to each training data set can be output, and the current residual sum of squares of multiple training data sets can be continuously calculated until the current residual sum of squares reaches the minimum or the current iteration number meets the conditions, that is, the iteration ends, and the final obtained parameter value is the optimal solution that makes the model predicted value closest to the real value. Among them, the final obtained parameter value is an array, which contains the optimal parameters obtained by fitting, and the code can extract the two optimal parameters, such as the optimal solution , where is the target first regression coefficient of the fiber cement-based material.

[0060] Further, the embodiment can also compare the real value (such as ) with the model predicted value (such as ) through the sklearn.metrics.r2_score function to obtain the value, where value = 1 - (residual sum of squares / total deviation sum of squares), that is, the proportion of variation explained by the model after deducting the error, The value of the value is in the range [0, 1], and the closer to 1 indicates that the model has a stronger ability to explain the data, and the predicted value is more consistent with the real value. In actual application, the value can be kept to 4 decimal places and printed to intuitively show the pros and cons of the model fitting, such as ​​The goodness-of-fit output can be 0.9039, indicating that the model can explain 90.39% of the data variation, and the fitting effect is good.

[0061] S103, constructing a first target prediction model of the fiber cement-based material according to the multiple sets of training data sets and the target first regression coefficient.

[0062] The first target prediction model is used to predict the actual second regression coefficient of the fiber cement-based material, and the actual second regression coefficient is used to calculate the actual compressive strength of the fiber cement-based material.

[0063] Although the above step fits a set of optimal solutions, the specific values still need to be further corrected. For example, this step can consider the dynamic influence of the cement proportion of the cementitious material, the sand-binder ratio, W / B, and fiber data (such as type, diameter, aspect ratio, and volume fraction) on the actual second regression coefficient, and dynamically correct the actual second regression coefficient by calibrating the target first regression coefficient, so that for each different mix proportion of the cement-based material, the target first regression coefficient is different on the basis of comprehensively considering the fiber-related factors and other factors.

[0064] Specifically, this step can comprehensively consider the training water-binder ratio, training cement strength, training cement ratio, training sand-binder ratio, and training fiber data of each set of training data sets, and the target first regression coefficient obtained in the previous step, to construct the first target prediction model of the fiber cement-based material. The method for constructing the first target prediction model is not limited, for example, a first preliminary prediction model can be selected according to actual needs, and the first preliminary prediction model is trained by using the multiple sets of training data sets and the target first regression coefficient to obtain the final first target prediction model. Alternatively, other methods can be used to construct the first target prediction model in this step, as long as the first target prediction model can be obtained.

[0065] S104, constructing a second target prediction model of the fiber cement-based material according to the multiple sets of training data sets and the first target prediction model.

[0066] The second target prediction model is used to predict the actual compressive strength of the fiber cement-based material.

[0067] Similar to step S103, the method for constructing the second target prediction model is not limited in this embodiment, for example, a second preliminary prediction model can be selected according to actual needs, and the second preliminary prediction model is trained by using the multiple sets of training data sets and the first target prediction model to obtain the final second target prediction model. Alternatively, other methods can be used to construct the second target prediction model in this step, as long as the second target prediction model can be obtained. ​​​​​

[0068] S105, calculate the mix proportion of the material incorporated in the fiber cement-based material based on the objective function and the constraint condition of the fiber cement-based material.

[0069] The objective function is determined based on the actual compressive strength predicted by the second target prediction model and the target compressive strength.

[0070] This step can calculate the mix proportion of the material incorporated in the fiber cement-based material by explicitly defining the objective function and the constraint condition of the fiber cement-based material. For example, different objective functions can be configured according to different design requirements of the fiber cement-based material. The number of configured objective functions is not limited. For example, a single objective function can be configured, and the objective function can aim to minimize the difference between the actual compressive strength predicted by the second target prediction model and the target compressive strength. Multiple objective functions can also be configured according to design requirements, and the mix proportion of the material incorporated in the fiber cement-based material can be calculated by balancing the trade-off relationship between multiple objective functions.

[0071] As a feasible implementation, the objective function includes a first objective function aiming to minimize the compressive strength error, a second objective function aiming to minimize the carbon emission, and a third objective function aiming to minimize the material cost. The first objective function is determined based on the actual compressive strength predicted by the second target prediction model and the target compressive strength.

[0072] The calculation of the mix proportion of the material incorporated in the fiber cement-based material based on the objective function and the constraint condition of the fiber cement-based material includes: determining the first function weight of the first objective function, the second function weight of the second objective function, and the third function weight of the third objective function, respectively; according to the first objective function, the second objective function, the third objective function, the first function weight, the second function weight, the third function weight, and the constraint condition, the mix proportion of the material incorporated in the fiber cement-based material is calculated by using a non-dominated sorting genetic algorithm, and the mix proportion at least includes the target cement strength, the target component ratio, the target water reducing agent proportion, the target water-binder ratio, the target sand-binder ratio, and the target fiber data.

[0073] Specifically, the first objective function is wherein, The actual compressive strength predicted by the second target prediction model can be The target compressive strength can be configured according to engineering requirements.

[0074] The target of the second objective function can be carbon emission minimization. For example, the second objective function can be based on the unit carbon emission coefficient of each component material (such as cement, mineral powder, fly ash, etc.), and the weighted cumulative addition is obtained to measure the impact of the fiber cement-based material ratio on the environment. For example, the second objective function can be .

[0075] The third objective function can calculate the total cost according to the usage of each material, which is used to reflect the economic objective, i.e. .

[0076] wherein the density Mi may be the usage of each raw material / fiber per unit volume in the fiber cement-based material, representing the mass of the i-th component per cubic meter of the fiber cement-based material, may represent the mass of carbon dioxide emitted by the i-th component per kilogram of the fiber cement-based material (i.e., the unit carbon emission coefficient), and the unit price may be the unit price of the i-th component in the fiber cement-based material, which may be further considered as the unit cost of each raw material / fiber based on the market price or the project price. Fibers may be a general term for fibers in the fiber cement-based material, which may include basalt fibers, PE fibers, etc. For example, as shown in Table 1, the characteristic parameters of the constituent materials in the fiber cement-based material are listed, including the density and the unit price P corresponding to each type of raw material / fiber.

[0077] Table 1 Characteristic parameters of constituent materials in fiber cement-based material

[0078]

[0079] Further, the constraint conditions can be configured according to actual conditions, which may include the proportion constraint condition and the range constraint condition of each constituent material of the fiber cement-based material, wherein the proportion constraint condition can be , which is used to indicate that the sum of the mass proportions of each cementitious material is close to 1 (the total amount of unit cementitious material), allowing a small error , such as a constant .

[0080] Table 2 can be the range constraint condition of each constituent material of the fiber cement-based material, OPC / B, FA / B, GGBFS / B and SF / B represent the mass proportions of ordinary portland cement, fly ash, slag powder and silica fume in the cementitious material, respectively, which helps to reflect the active component structure of the cementitious material; SP / B represents the mass ratio of water reducing agent to cementitious material; S / B represents the mass ratio of sand to cementitious material, W / B represents the water-binder ratio, Vf represents the fiber volume fraction, and df and Lf represent the diameter and length of the fiber, respectively.

[0081] Table 2 Range constraint condition

[0082]

[0083] In specific applications, the corresponding weights can be configured for different objective functions according to actual project requirements. For example, in the construction of super high-rise building core tube, structural safety is the primary criterion, and the compressive strength needs to reach C80 and above. At this time, the engineer can increase the compressive strength weight to 0.6, and set the carbon footprint and cost weights to 0.2 and 0.2 respectively. Then the algorithm will prioritize ensuring that the strength predicted by the second target prediction model meets the standard, and on this basis, it will try to reduce carbon emissions and cost as much as possible. The final generated scheme can mainly use high-grade cement, supplemented by a small amount of high-performance mineral admixture, to ensure that the strength meets the load requirements of super high-rise buildings, while controlling the increase of carbon footprint through reasonable collocation of mineral admixtures, and avoiding excessive rise in cost.

[0084] For example, for municipal road base engineering, the project has relatively loose requirements on strength (such as C30), but due to the large amount of engineering and sensitive total carbon emissions, and the need to strictly control the project cost. The engineer can set the carbon footprint weight to 0.5, the cost weight to 0.4, and the compressive strength weight to only 0.1. At this time, the algorithm will focus on screening mix proportion schemes with low cement dosage (such as a large amount of industrial waste such as fly ash and slag), sacrificing a small amount of non-critical strength margin, significantly reducing cement dosage (cement is the main source of carbon footprint), while using low-cost admixtures to control material cost, and the final generated scheme can present the characteristics of "medium strength + low carbon emission + low cost" to adapt to the economic and environmental needs of municipal engineering.

[0085] For example, in green building demonstration projects, the carbon footprint index is included in the compulsory assessment system, but if the project is located in a high earthquake-prone area, the structural strength still needs to reach a high level (such as C40). The engineer can set the compressive strength weight to 0.4, the carbon footprint weight to 0.4, and the cost weight to 0.2. At this time, the algorithm will prioritize selecting cementitious material combinations with lower carbon emissions (such as using low-carbon cement + metakaolin composite system) on the premise of ensuring that the strength is not lower than the design value, while balancing the premium of low-carbon materials in the material procurement link, and the generated scheme not only meets the carbon emission reduction requirements of green buildings, but also ensures structural safety, and the cost is slightly increased but controlled within the acceptable range of green technology increment of the project.

[0086] Specifically, the present embodiment can use a non-dominated sorting genetic algorithm to first randomly generate an initial parent population according to the design parameter range; then the compressive strength predicted by the second target prediction model is used as the fitness function of the non-dominated sorting genetic algorithm, and the carbon footprint and cost are used as the constraint conditions of the non-dominated sorting genetic algorithm. The carbon footprint calculation formula and the material cost calculation formula are taken as objective functions, and the population is subjected to non-dominated sorting and congestion calculation in combination with the set constraint conditions; the offspring population is generated through selection, crossover and mutation operations, and iteration is repeated until the maximum evolution number is reached, and finally a plurality of Pareto optimal solutions are obtained, reflecting the trade-off relationship between the objectives, such as the Technique for Order Preference by Similarity to an Ideal Solution (TOPSIS), and the weight of the compressive strength, the weight of the carbon footprint and the weight of the cost are adjusted as needed, and the final scheme is selected from the plurality of Pareto optimal solutions.

[0087] The final scheme is the mixing proportion of the material in the fiber cement-based material, including the cement strength , the mass ratio of ordinary Portland cement, fly ash, slag powder and silica fume in the cementitious material OPB / B, FA / B, GGBFS / B and SF / B (i.e. target component ratio), the mass ratio of sand to cementitious material S / B (i.e. target sand-binder ratio), the mass ratio of water reducing agent to cementitious material SP / B (i.e. target water reducing agent proportion), the water-binder ratio W / B (i.e. target water-binder ratio), and the detailed properties of the fiber (i.e. target fiber data), such as fiber type, diameter df, length Lf, aspect ratio, density, tensile strength, elastic modulus, percentage, etc.

[0088] The fiber data is integrated into the compressive strength formula to determine two regression coefficients suitable for fiber cement-based materials, and then a prediction model for predicting the compressive strength of the fiber cement-based material is constructed by using the determined two regression coefficients, so that the influence of fiber incorporation is considered when calculating the material mixing proportion, the design precision of the mixing proportion is improved, and the precise regulation and control of the performance of the fiber cement-based material is realized.

[0089] Figure 2 is a flowchart of a material proportioning intelligent optimization method combining field knowledge and machine learning provided by another embodiment of the present application. Each set of training data set of the present embodiment further includes: training water-binder ratio, training cement strength, training compressive strength, and the compressive strength formula is , wherein, represents the compressive strength, represents the cement strength, W / B represents the water-binder ratio, respectively represent the first regression coefficient and the second regression coefficient; and the first target prediction model of the fiber cement-based material is further optimized according to the multiple sets of training data sets and the target first regression coefficient, that is, for each set of training data set, the training water-binder ratio, the training cement strength, the training compressive strength and the target first regression coefficient of each set of training data set are input into the compressive strength formula to obtain the theoretical second regression coefficient corresponding to each set of training data set; a first preliminary prediction model of the fiber cement-based material is established, and the first preliminary prediction model is a random forest regressor; and the first preliminary prediction model is trained according to the training water-binder ratio, the training cement strength, the training cement ratio, the training sand-binder ratio and the training fiber data of the multiple sets of training data sets, and the theoretical second regression coefficient corresponding to each set of training data set, to obtain the first target prediction model of the fiber cement-based material. As shown in Figure 2 , the method comprises:

[0090] S201, a plurality of sets of training data sets of a fiber cement-based material are obtained.

[0091] S202, a target first regression coefficient of the fiber cement-based material is determined based on the multiple sets of training data sets and a compressive strength formula of the cement-based material.

[0092] S203, for each set of training data set, the training water-binder ratio, the training cement strength, the training compressive strength and the target first regression coefficient of each set of training data set are input into the compressive strength formula to obtain the theoretical second regression coefficient corresponding to each set of training data set.

[0093] Specifically, each set of training data set collected in the above step can include the proportion of cement in cementitious material (i.e., the training cement ratio), the sand-binder ratio (i.e., the training sand-binder ratio), the W / B (i.e., the training water-binder ratio), the fiber type, the fiber diameter, the fiber length, the fiber aspect ratio, the fiber volume percentage (i.e., the training fiber data), the (i.e., the training cement strength), (i.e., the training compressive strength), etc. The W / B, , of each set of training data set can be extracted in this step, and = 1.61 (i.e., the target first regression coefficient), and then the theoretical second regression coefficient corresponding to each set of training data set can be calculated according to the compressive strength formula, that is, .

[0094] S204, a first preliminary prediction model of the fiber cement-based material is established, and the first preliminary prediction model is a random forest regressor.

[0095] This step can establish a random forest regressor of the fiber cement-based material according to the above data set. The prediction model (i.e., the first preliminary prediction model) can preliminarily determine the input features of the first preliminary prediction model as the proportion of cement in cementitious materials, the sand-binder ratio, W / B, the type of fiber, the fiber diameter, the fiber length, the fiber aspect ratio, the fiber volume percentage, , and the output feature as the theoretical second regression coefficient .

[0096] Further, before model training, the input features can be preprocessed, such as the input features and output features can be spliced into a feature matrix in the form of a column vector in advance, the correlation coefficient matrix between the input feature columns (excluding the last column) is calculated, the correlation between the features is visualized through a heatmap (seaborn.heatmap), which facilitates preliminary understanding of the variable relationship; the classification feature FT (i.e., the type of fiber) is one-hot encoded (pd.get_dummies), which is converted into a numerical variable that can be processed by the model, and the output feature column is again confirmed to exist in the data.

[0097] S205, according to the training water-binder ratio, the training cement strength, the training cement ratio, the training sand-binder ratio and the training fiber data of the plurality of training data sets, and the corresponding theoretical second regression coefficient of each training data set, the first preliminary prediction model is trained to obtain the first target prediction model of the fiber cement-based material.

[0098] In the specific model training process, first, the input feature data set and the output feature data can be divided into a training set (X_train, Y_train) and a test set (X_test, Y_test) according to a 9:1 ratio, to ensure the objectivity of model evaluation.

[0099] Then, the random forest regressor (RandomForestRegressor) can be initialized, and hyperparameter tuning can be performed through grid search (GridSearchCV). The parameter grid can include key parameters such as the number of trees (n_estimators), the maximum depth (max_depth), and the minimum number of samples for splitting and leaf nodes, 5-fold cross-validation (cv=5) is adopted, the score is used as the evaluation index, and the optimal parameter combination is selected.

[0100] Finally, the random forest model with the optimal parameters (i.e., the first target prediction model) can be used to make predictions on the training set and the test set, respectively, and the mean square error (MSE) and the goodness of fit of the training set and the test set are calculated and output, to evaluate the fitting effect and generalization ability of the model; the model stability is further verified through 5-fold cross-validation, the average score and the standard deviation of cross-validation are output, to comprehensively reflect the performance of the model.​

[0101] Figure 3 This application provides a second regression coefficient in one embodiment. A diagram showing the comparison between the actual and predicted values, such as... Figure 3 As shown, the horizontal axis represents the second regression coefficient. The true value is shown on the left, and the predicted value of the second regression coefficient output by the first target prediction model is shown on the right. In this embodiment, the prediction results of the first target prediction model on the training set and the test set are calculated respectively, and the prediction results are compared with their respective true values. The scores were used as evaluation metrics for the model. Specifically, the evaluation metrics for the first objective prediction model on the training and test sets were as follows: =0.991 and =0.949.

[0102] S206. Based on multiple training datasets and the first target prediction model, construct a second target prediction model for fiber cement-based materials.

[0103] S207. Based on the objective function and constraints of fiber cementitious materials, calculate the mix proportion of admixtures in fiber cementitious materials.

[0104] This embodiment provides a material proportioning intelligent optimization method that integrates domain knowledge and machine learning. By calculating the theoretical second regression coefficient corresponding to each training dataset, and then training the first preliminary prediction model based on the training water-cement ratio, training cement strength, training cement ratio, training mortar-cement ratio, and training fiber data of multiple training datasets, as well as the theoretical second regression coefficient corresponding to each training dataset, a first target prediction model for fiber cementitious materials is obtained. This improves the accuracy of the first target prediction model and provides an accurate model basis for subsequent calculation of the proportion of admixtures in fiber cementitious materials, further realizing precise control of the performance of fiber cementitious materials.

[0105] As a feasible implementation method, a second target prediction model for fiber cementitious materials is constructed based on multiple training datasets and the first target prediction model, including:

[0106] For each training dataset, the training water-cement ratio, training cement strength, training cement ratio, training sand-cement ratio, and training fiber data of each training dataset are input into the first target prediction model to obtain the actual second regression coefficients corresponding to each training dataset.

[0107] A second preliminary prediction model for fiber cement-based materials is established. The second preliminary prediction model is a multilayer perceptron model with two hidden layers, each containing 64 neurons, and the activation function of the multilayer perceptron model is the ReLU function.

[0108] Based on each training dataset and the corresponding actual second regression coefficients, the second preliminary prediction model is trained to obtain the second target prediction model for fiber cement-based materials.

[0109] In the specific process of constructing the second target prediction model, the training data and benchmark model required for model training can be prepared in advance. For example, the data for prediction can be extracted from each training dataset. The feature data (i.e., the second regression coefficients), such as training water-cement ratio, training cement strength, training cement ratio, training mortar-cement ratio, and training fiber data, can be input into the first target prediction model to obtain the actual second regression coefficients corresponding to each training dataset. Further, a Multilayer Perceptron (MLP) model (i.e., the second preliminary prediction model) can be defined. Specifically, an MLPModel class can be defined, inheriting from nn.Module, to construct a multilayer perceptron model with two hidden layers (64 neurons per layer) and a ReLU activation function. The input dimension can be determined based on the feature dimension of the training set, and the output dimension is 1, used to predict the 28-day compressive strength of fiber-reinforced cementitious materials.

[0110] Subsequently, based on each training dataset and the corresponding actual second regression coefficients, the second preliminary prediction model is trained to obtain the second target prediction model for fiber cementitious materials. The model training process can be as follows: first, the theoretical compressive strength is calculated based on the actual second regression coefficients of each training dataset; second, the predicted compressive strength output by the second preliminary prediction model is calculated based on the specific content of each training dataset; and third, the second preliminary prediction model is gradually trained so that the output predicted compressive strength is close to the theoretical compressive strength, that is, the error between the predicted compressive strength and the theoretical compressive strength is gradually reduced. When the number of training iterations reaches the maximum, the final second target prediction model can be obtained, which can be used to predict the actual compressive strength of fiber cementitious materials.

[0111] As a feasible implementation method, each training dataset also includes: the ratio of training components and the proportion of water-reducing agent in the training dataset;

[0112] Based on each training dataset and the corresponding actual second regression coefficients, the second preliminary prediction model is trained to obtain the second target prediction model for fiber cementitious materials, including:

[0113] The second preliminary prediction model was determined to be the second current prediction model for fiber cementitious materials;

[0114] inputting the training component ratio, the training water reducing agent proportion, the training water-binder ratio, the training sand-binder ratio, the training fiber data and the training cement strength of each training data set into the second current prediction model to obtain the predicted compressive strength corresponding to each training data set;

[0115] inputting the training water-binder ratio, the training cement strength, the target first regression coefficient and the actual second regression coefficient of each training data set into the compressive strength formula to obtain the theoretical compressive strength corresponding to each training data set;

[0116] updating the model parameters of the second current prediction model according to the predicted compressive strength and the theoretical compressive strength corresponding to each training data set;

[0117] returning to the step of inputting the training component ratio, the training water reducing agent proportion, the training water-binder ratio, the training sand-binder ratio, the training fiber data and the training cement strength of each training data set into the second current prediction model to obtain the predicted compressive strength corresponding to each training data set, until the updating number of the second current prediction model reaches the maximum iteration number;

[0118] determining the second current prediction model corresponding to the maximum iteration number as the second target prediction model of the fiber cement-based material.

[0119] In the specific process of each training round, the model parameters of the second current prediction model can be updated gradually according to the predicted compressive strength and the theoretical compressive strength corresponding to each training data set. The specific process of updating the model parameters is not limited, for example, the model parameters of the second current prediction model can be updated directly according to the size between the predicted compressive strength and the theoretical compressive strength, or the model parameters of the second current prediction model can be updated according to a series of loss calculations on the predicted compressive strength and the theoretical compressive strength, or the model parameters of the second current prediction model can be updated according to other ways, and the present embodiment does not make further limitation.

[0120] As a feasible implementation manner, updating the model parameters of the second current prediction model according to the predicted compressive strength and the theoretical compressive strength corresponding to each training data set comprises:

[0121] calculating the physical loss value and the data loss value of each training data set based on the predicted compressive strength and the theoretical compressive strength corresponding to each training data set, respectively;

[0122] calculating the total loss value of the multiple training data sets based on the physical loss value and the data loss value of each training data set;

[0123] updating the model parameters of the second current prediction model according to the total loss value.

[0124] In the updating process of specific model parameters, the updating of the model parameters can be comprehensively guided by calculating the physical loss value and the data loss value in each training round, and the training of the second current prediction model is gradually realized. The physical loss value can reflect the physical consistency loss of the plurality of training data sets, and the model output can be matched with the observation data and the dynamic behavior of the physical system according to the physical consistency loss, such as the physical loss value can be the error between the predicted compressive strength and the theoretical compressive strength; the data loss value can reflect the supervised learning loss of the plurality of training data sets, so as to optimize the model parameters to reduce the error according to the difference between the model prediction result and the real label, such as the data loss value can be the error between the predicted compressive strength and the real strength.

[0125] The main process of constructing the second target prediction model is described below:

[0126] (1) First, a multilayer perceptron (MLP) model can be selected as the main model, and the training data of the MLP model is read and preprocessed, for example, the input features of the MLP model are extracted from the plurality of training data sets: the proportion of various components of the fiber cement-based material, including cement, fly ash, mineral powder, silica fume, etc. Cementitious material, sand cement ratio, water cement ratio, high efficiency water reducing agent percentage, cement strength , and detailed properties of the fiber (fiber type, diameter, length, aspect ratio, density, tensile strength, elastic modulus, percentage), and the output feature can be the 28-day compressive strength (i.e. target variable); Then the training data can be divided into a feature matrix X_raw (composed of input features, excluding columns) and a target variable Y_raw (column value); The feature matrix is further distinguished between numerical and categorical types, and the categorical features (FT) are one-hot encoded by OneHotEncoder. The encoded categorical features and numerical features can be spliced to obtain a complete feature matrix X_all. The above preliminarily processed data is divided into a training set and a test set in a ratio of 9:1, and the feature matrix and the target variable are standardized to have zero mean and unit variance, which facilitates subsequent model training.

[0127] (2) Then, the random forest related variables can be prepared, and the specific steps can include:

[0128] 1) First, there is a trained random forest model file "rf_alpha_a.pkl" (i.e. the first target prediction model), which can be loaded into the code by the "joblib.load" function to obtain the model object "rf_alpha_a".

[0129] 2) The feature names that can be explicitly used for prediction , i.e. C represents the proportion of cement in cementitious materials, SCR represents the sand-cement ratio, W / B represents the water-binder ratio, FDia represents the fiber diameter, FL represents the fiber length, FAR represents the fiber aspect ratio, FV represents the fiber volume fraction, represents the cement strength, FT_BF represents BF fiber, FT_PE represents PE fiber, FT_POM represents POM fiber, FT_PP represents PP fiber, FT_PVA represents PVA fiber, FT_carbon fiber represents carbon fiber, the above feature names are stored in the "alpha_a_features" list, and need to correspond to the features used when training the random forest model.

[0130] 3) The code determines the index position of the above features in all preprocessed features. Because the features include numerical and category types after one-hot encoding, first record the number of numerical features "n_numerical", then traverse each feature in "alpha_a_features", if the feature is in the numerical feature column name "numerical_cols", add its index in "numerical_cols" to the "alpha_a_feature_indices" list; if not, traverse the category feature name after one-hot encoding "cat_feature_names", when the name starting with the feature is found, add its corresponding index (based on the number of numerical features plus the current index position) to the "alpha_a_feature_indices" list.

[0131] 4) Based on the column names of all features "X_feature_names" (consisting of numerical feature column names and category feature names after one-hot encoding), through the index recorded in "alpha_a_feature_indices", the input feature column names for random forest model prediction can be obtained and stored in the "rf_feature_names" list, which is ready for subsequent prediction using the random forest model and combined with physical model calculation and other operations.

[0132] (3) Then, the dataset and data loader can be constructed: define the ConcreteDataset class, which inherits from Dataset, convert the input features and target variable to tensors in the class, and implement methods to get the length of the dataset and get data samples by index, such as creating an instance of the class based on the training dataset train_dataset, and using DataLoader to encapsulate it, setting the batch size to 32 and shuffling the data order to facilitate loading data by batch during training.

[0133] (4) Then, the physical loss function can be defined and used to calculate the physical loss: define a physical loss function named "physics_loss", which introduces the physical theoretical law of fiber cement-based material strength into model training to constrain the prediction results of the neural network to comply with domain knowledge.

[0134] In the process of calculating the physical loss, first, the physics_loss function can convert the input feature tensor to a NumPy array through "detach().cpu().numpy()", which is to adapt to the input format of the subsequent random forest model. Then, the features used to predict can be extracted from the converted array through "alpha_a_feature_indices" indexing, and these feature data can be organized into a DataFrame with column names corresponding to "rf_feature_names" used during random forest training to ensure that the input format matches.

[0135] Subsequently, the pre-trained random forest model file can be used to predict the above features to obtain the key parameters in the physical formula. At the same time, the values of cement strength ( ) and water-binder ratio (W / B) are located and extracted from the input features, and adjusted to column vector form to meet the subsequent calculation requirements.

[0136] In order to realize tensor operation, the predicted by the random forest model file can be converted to a PyTorch tensor, and the shape is adjusted and transferred to the same device (GPU or CPU) as the input features. Then, the theoretical strength can be calculated according to the compressive strength formula of cement-based materials, where is a fixed parameter set based on domain knowledge (default 1.61).

[0137] Finally, the model prediction strength and the theoretical strength The difference between them is obtained and returned as a physical loss.

[0138] (5) Finally, the trained model can be combined with the physical loss: first, the code will automatically select the training device according to the hardware conditions, and prefer to use GPU to improve the calculation efficiency, if there is no GPU, then use CPU; Then the multi-layer perception (MLP) model can be initialized, and the parameters are migrated to the selected device, and the Adam optimizer (learning rate is 0.001) and the mean square error loss function (MSELoss) can be configured for subsequent parameter optimization and data loss calculation. Before the training process starts, 100 training rounds (such as num_epochs=100) can be set in advance, and the weight of data loss and physical loss can be balanced by setting lambda_, wherein the optimal training parameter lambda_=0.0064 can be obtained according to manual parameter tuning.

[0139] In each training round, the model can be switched to the training mode, and the cumulative variables of the total loss, data loss and physical loss can be initialized; the data can be read in batches through the training data loader (train_loader), and each batch of data can contain 32 samples, and the input features and real labels are migrated to the training device.

[0140] For each batch of data, the model can predict the strength of the fiber cement-based material by predicting the input features, and then the total loss can be calculated by calculating the data loss and the physical loss respectively: wherein the data loss (loss_data) can be the mean square error between the predicted value and the real label, to ensure that the model fits the training data; The physical loss (loss_phys) can be calculated by calling the defined physics_loss function, which constrains the model prediction to comply with the theoretical law of the strength of the fiber cement-based material; the total loss is the sum of the data loss and the physical loss multiplied by the weight, such as .

[0141] After the loss calculation is completed, the previous gradient is cleared through the optimizer (optimizer.zero_grad()), the total loss is calculated by back propagation (loss.backward()), and the model parameters of the multi-layer perception model are updated according to the gradient (optimizer.step()). At the same time, the code will accumulate the loss value of each batch, and print the current total loss, data loss and physical loss every 10 rounds, which is convenient for monitoring the training progress and loss trend.

[0142] After 100 training rounds, the code will output a "Training Complete" message. The entire training process combines data-driven and physical constraint-guided model training, ensuring both the model's ability to fit the data, making the predictions closely match the actual data, and ensuring that the predictions conform to the physical laws of fiber cementitious materials, thereby improving the model's generalization ability and the rationality of its predictions.

[0143] Furthermore, after model training is complete, the model can be set to evaluation mode, and predictions can be made on both the test and training sets. During prediction, the data can first be converted into tensors and transferred to the device. After the model outputs the results, they can be converted back into NumPy arrays, and the actual predicted values ​​can be obtained through inverse standardization transformation. Then, the coefficient of determination can be calculated using R2_score. To evaluate the model's goodness of fit to the data, the mean absolute error (MAE) is calculated using Mean_Absolute_Error to measure the average deviation between the predicted and actual values. Finally, the evaluation results on the training and test sets can be printed.

[0144] Figure 4 This is one embodiment provided in this application. A diagram showing the comparison between the actual and predicted values, such as... Figure 4 As shown in (a) and (b), the outputs of different second-objective prediction models were compared for the training and test sets, respectively. The difference between the predicted value and their respective true value, and in terms of The score was used as an evaluation metric for model assessment, among which, Figure 4 The second target prediction model in (a) is a neural network (NN) model that does not consider physical constraints; that is, the predicted value in (a) is the result of the compressive strength neural network prediction without incorporating physical formulas. Figure 4 The second target prediction model in (b) is a Physics-Informed Neural Network (PINN) model that considers physical constraints. That is, the predicted value in (b) is the result of a compressive strength neural network incorporating physical formulas. It can be seen that, regardless of whether it is for the training set or the test set, Figure 4 The prediction performance of (b) is improved compared to (a) in all aspects, such as the evaluation metrics on the training set. The score improved from 0.894 to 0.966 in the test set evaluation metrics. The PINN model improved from 0.839 to 0.925, showing a better fit to the data than the NN model, and the error between the predicted intensity and the actual intensity was smaller.

[0145] As can be seen from the above description, the material proportioning intelligent optimization method provided by the embodiment fuses field knowledge and machine learning, involves a double-drive design fusing field professional knowledge of cement-based materials and machine learning algorithms, is suitable for mix proportion design, performance prediction and multi-objective optimization of fiber-reinforced cement-based materials, can be widely applied to fields such as building structures, bridge engineering and municipal construction that have comprehensive needs for material ductility, strength, economy and low carbon, and provides technical support for research and application of high-performance and green cement-based materials.

[0146] For example, the embodiment constructs a flexible and extensible multi-objective optimization framework instead of being limited to a single fixed scheme. For engineers, the optimization direction can be dynamically adjusted according to the core needs of specific projects in actual engineering scenarios, such as fine-tuning the weights of compressive strength, carbon footprint and cost to quickly generate multiple groups of mix proportion schemes adapted to different priorities, and finally find the balance point most suitable for the actual project among performance compliance, green low carbon and economic controllability, providing multiple selectable schemes for engineers in specific design.

[0147] Whether it is a special structure that focuses on strength guarantee, a people's livelihood project that pursues the best cost performance, or a benchmark project that focuses on low carbon, the weight adjustment can quickly lock the adaptive proportioning direction, and simultaneously output specific parameters (such as material usage, predicted strength value, carbon footprint value, cost details, etc.) corresponding to multiple Pareto optimal solutions, allowing engineers to balance other dimensions while grasping the core needs in scheme selection, significantly improving the efficiency and accuracy of mix proportion design.

[0148] Corresponding to the material proportioning intelligent optimization method fusing field knowledge and machine learning of the above embodiment, Figure 5 is a structural diagram of a material proportioning intelligent optimization device fusing field knowledge and machine learning according to an embodiment of the present application. For ease of illustration, only parts related to the embodiments of the present application are shown.

[0149] With reference to Figure 5 , the device comprises:

[0150] The acquisition module 301 is configured to acquire multiple sets of training data sets of fiber cement-based materials, wherein each set of training data sets at least includes training cement ratio, training sand-binder ratio and training fiber data of the fiber cement-based materials.

[0151] The regression coefficient determination module 302 is configured to determine a target first regression coefficient of the fiber cement-based materials based on the multiple sets of training data sets and a compressive strength formula of the cement-based materials, wherein the compressive strength formula is determined based on at least a first regression coefficient and a second regression coefficient of the cement-based materials, and the target first regression coefficient is used to calculate the actual compressive strength of the fiber cement-based materials.

[0152] The first model construction module 303 is configured to construct a first target prediction model of the fiber cement-based material according to the multiple sets of training data sets and the target first regression coefficient, the first target prediction model being used to predict an actual second regression coefficient of the fiber cement-based material, the actual second regression coefficient being used to calculate an actual compressive strength of the fiber cement-based material.

[0153] The second model construction module 304 is configured to construct a second target prediction model of the fiber cement-based material according to the multiple sets of training data sets and the first target prediction model, the second target prediction model being used to predict an actual compressive strength of the fiber cement-based material.

[0154] The calculation module 305 is configured to calculate a mix proportion of the material incorporated in the fiber cement-based material based on a target function and a constraint condition of the fiber cement-based material, wherein at least part of the target function is determined based on the actual compressive strength output by the second target prediction model and a target compressive strength.

[0155] The device for intelligently optimizing material proportioning by fusing field knowledge and machine learning provided in the embodiment comprises a obtaining module configured to obtain multiple sets of training data sets of a fiber cement-based material, wherein each set of training data sets at least comprises training cement ratios, training sand-binder ratios and training fiber data of the fiber cement-based material; a regression coefficient determination module configured to determine a target first regression coefficient of the fiber cement-based material based on the multiple sets of training data sets and a compressive strength formula of the cement-based material, the compressive strength formula being determined based on at least a first regression coefficient and a second regression coefficient of the cement-based material, the target first regression coefficient being used to calculate an actual compressive strength of the fiber cement-based material; a first model construction module configured to construct a first target prediction model of the fiber cement-based material according to the multiple sets of training data sets and the target first regression coefficient, the first target prediction model being used to predict an actual second regression coefficient of the fiber cement-based material, the actual second regression coefficient being used to calculate the actual compressive strength of the fiber cement-based material; a second model construction module configured to construct a second target prediction model of the fiber cement-based material according to the multiple sets of training data sets and the first target prediction model, the second target prediction model being used to predict the actual compressive strength of the fiber cement-based material; and a calculation module configured to calculate a mix proportion of the material incorporated in the fiber cement-based material based on a target function and a constraint condition of the fiber cement-based material, wherein at least part of the target function is determined based on the second target prediction model and a target compressive strength. By fusing the fiber data into the compressive strength formula, two regression coefficients suitable for the fiber cement-based material are determined, and then by using the two determined regression coefficients to construct a prediction model for predicting the compressive strength of the fiber cement-based material, the influence of the fiber incorporation is considered when calculating the material mix proportion, the design precision of the mix proportion is improved, and thus the performance of the fiber cement-based material is accurately regulated and controlled.

[0156] Optionally, each training data set further comprises: training water-binder ratio, training cement strength, training compressive strength, and the compressive strength formula is wherein, represents the compressive strength, represents the cement strength, W / B represents the water-binder ratio, respectively represent the first regression coefficient and the second regression coefficient;

[0157] The first model construction module is specifically configured to:

[0158] For each training data set, the training water-binder ratio, the training cement strength, the training compressive strength, and the target first regression coefficient of each training data set are input into the compressive strength formula to obtain a theoretical second regression coefficient corresponding to each training data set.

[0159] The first preliminary prediction model of the fiber cement-based material is a random forest regressor.

[0160] According to the training water-binder ratio, the training cement strength, the training cement ratio, the training sand-binder ratio, and the training fiber data of the plurality of training data sets, and the theoretical second regression coefficient corresponding to each training data set, the first preliminary prediction model is model trained to obtain the first target prediction model of the fiber cement-based material.

[0161] Optionally, the second model construction module comprises:

[0162] The input unit is configured to, for each training data set, input the training water-binder ratio, the training cement strength, the training cement ratio, the training sand-binder ratio, and the training fiber data of each training data set into the first target prediction model to obtain an actual second regression coefficient corresponding to each training data set.

[0163] The model establishment unit is configured to establish a second preliminary prediction model of the fiber cement-based material, wherein the second preliminary prediction model is a multi-layer perceptron model, the multi-layer perceptron model has two hidden layers, each hidden layer contains 64 neurons, and the activation function of the multi-layer perceptron model is a ReLU function.

[0164] The model training unit is configured to, according to each training data set and the actual second regression coefficient corresponding to each training data set, model train the second preliminary prediction model to obtain the second target prediction model of the fiber cement-based material.

[0165] Optionally, each training data set further comprises: training component ratio and training water-reducing agent proportion;

[0166] The model training unit comprises:

[0167] The first determining sub-unit is configured to determine the second preliminary prediction model as a second current prediction model of the fiber cement-based material.

[0168] The first input sub-unit is configured to input the training component ratio, the training water reducing agent ratio, the training water-binder ratio, the training sand-binder ratio, the training fiber data, and the training cement strength of each training data set into the second current prediction model to obtain the predicted compressive strength corresponding to each training data set.

[0169] The second input sub-unit is configured to input the training water-binder ratio, the training cement strength, the target first regression coefficient, and the actual second regression coefficient of each training data set into the compressive strength formula to obtain the theoretical compressive strength corresponding to each training data set.

[0170] The updating sub-unit is configured to update the model parameters of the second current prediction model according to the predicted compressive strength and the theoretical compressive strength corresponding to each training data set.

[0171] The returning execution sub-unit is configured to return to execute the step of inputting the training component ratio, the training water reducing agent ratio, the training water-binder ratio, the training sand-binder ratio, the training fiber data, and the training cement strength of each training data set into the second current prediction model to obtain the predicted compressive strength corresponding to each training data set until the number of updates of the second current prediction model reaches the maximum iteration number.

[0172] The second determining sub-unit is configured to determine the second current prediction model corresponding to the maximum iteration number as a second target prediction model of the fiber cement-based material.

[0173] Optionally, the updating sub-unit is specifically configured to:

[0174] Based on the predicted compressive strength and the theoretical compressive strength corresponding to each training data set, the physical loss value and the data loss value of each training data set are calculated respectively.

[0175] Based on the physical loss value and the data loss value of each training data set, the total loss value of the multiple training data sets is calculated.

[0176] The model parameters of the second current prediction model are updated according to the total loss value.

[0177] Optionally, the regression coefficient determining module is specifically configured to:

[0178] The first initial value is determined as the current first regression coefficient of the fiber cement-based material, and the second initial value is determined as the current second regression coefficient of the fiber cement-based material.

[0179] inputting the training water-binder ratio, the training cement strength, the current first regression coefficient and the current second regression coefficient of each training data set into the compressive strength formula to obtain the theoretical compressive strength of each training data set;

[0180] based on the theoretical compressive strength and the training compressive strength of each training data set, calculating the current residual sum of squares of the plurality of training data sets;

[0181] According to the current residual sum of squares, the current first regression coefficient and the current second regression coefficient are modified, and the step of inputting the training water-binder ratio, the training cement strength, the current first regression coefficient and the current second regression coefficient of each training data set into the compressive strength formula to obtain the theoretical compressive strength of each training data set is returned until the current residual sum of squares reaches convergence or the number of modifications of the current first regression coefficient reaches the maximum iteration number.

[0182] The current first regression coefficient corresponding to the current residual sum of squares reaching convergence or the number of modifications of the current first regression coefficient reaching the maximum iteration number is determined as the target first regression coefficient of the fiber cement-based material.

[0183] Optionally, the objective function includes a first objective function aiming to minimize the compressive strength error, a second objective function aiming to minimize the carbon emission, and a third objective function aiming to minimize the material cost, and the first objective function is determined based on the actual compressive strength predicted by the second target prediction model and the target compressive strength.

[0184] The calculation module is specifically configured to:

[0185] determine a first function weight of the first objective function, a second function weight of the second objective function, and a third function weight of the third objective function, respectively;

[0186] According to the first objective function, the second objective function, the third objective function, the first function weight, the second function weight, the third function weight, and the constraint condition, the non-dominated sorting genetic algorithm is used to calculate the mix proportion of the mixed material in the fiber cement-based material, and the mix proportion at least includes the target cement strength, the target component ratio, the target water reducer proportion, the target water-binder ratio, the target sand-binder ratio and the target fiber data.

[0187] It should be noted that the information interaction, execution process and the like between the above-mentioned devices / units are based on the same concept as the method embodiments of the present application, and the specific functions and technical effects brought about can be referred to the method embodiments part, which will not be repeated here.

[0188] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is merely an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiments can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit. Furthermore, the specific names of the functional units and modules are only for easy differentiation and are not intended to limit the scope of protection of this application. The specific working process of the units and modules in the above system can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0189] Figure 6 This is a schematic diagram of the structure of a terminal device provided in one embodiment of this application, as shown below. Figure 6 As shown, the terminal device 500 of this embodiment includes: at least one processor 502 ( Figure 6 (Only one is shown) a processor, a memory 501, and a computer program 503 stored in the memory 501 and executable on at least one processor 502. When the processor 502 executes the computer program 503, it implements the steps in the control method embodiments of any of the above-described application programs.

[0190] Terminal device 500 can be a computing device such as a desktop computer, laptop, handheld computer, or cloud server. This terminal device may include, but is not limited to, a processor 502 and a memory 501. Those skilled in the art will understand that... Figure 6 This is merely an example of terminal device 500 and does not constitute a limitation on terminal device 500. It may include more or fewer components than shown in the figure, or combine certain components, or different components, such as input / output devices, network access devices, etc.

[0191] The processor 502 can be a central processing unit (CPU), and can also be other general-purpose processors, a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field-programmable gate array (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor.

[0192] The memory 501 can be an internal storage unit of the terminal device 500 in some embodiments, for example, a hard disk or a memory of the terminal device 500. The memory 501 can also be an external storage device of the terminal device 500 in other embodiments, for example, a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, etc. equipped on the terminal device 500. Further, the memory 501 can include both the internal storage unit and the external storage device of the terminal device 500. The memory 501 is used to store an operating system, application programs, a boot loader, data and other programs, for example, program codes of computer programs, etc. The memory 501 can also be used to temporarily store data that has been output or will be output.

[0193] The embodiments of the present application further provide a computer readable storage medium, which stores a computer program. The computer program is executed by the processor 502 to implement the steps in the above-mentioned various method embodiments.

[0194] The embodiments of the present application provide a computer program product. When the computer program product is run on a terminal device, the terminal device can implement the steps in the above-mentioned various method embodiments.

[0195] The integrated unit, if implemented in the form of a software function unit and sold or used as an independent product, can be stored in a computer readable storage medium. Based on such understanding, the present application can implement all or part of the processes in the above-mentioned embodiment methods through a computer program to instruct relevant hardware to complete, and the computer program can be stored in a computer readable storage medium. When the computer program is executed by the processor 502, the steps of each method embodiment described above can be implemented. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or some intermediate forms. The computer readable storage medium at least includes any entity or device capable of carrying the computer program code to the apparatus / terminal device, recording medium, computer memory, read-only memory (Read-Only Memory, ROM), random access memory (Random Access Memory, RAM), electrical carrier signal, telecommunication signal, and software distribution medium. For example, U disk, mobile hard disk, magnetic disk or optical disk, etc. In some jurisdictions, according to legislation and patent practice, the computer readable storage medium can not be an electrical carrier signal and a telecommunication signal.

[0196] In the above embodiments, the description of each embodiment has its own focus, and the parts not described or recorded in detail in a certain embodiment can be referred to the related description of other embodiments.

[0197] Those of ordinary skill in the art can realize that the units and algorithm steps of each example described in combination with the embodiments disclosed herein can be realized in electronic hardware or a combination of computer software and electronic hardware. Whether the functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. A person skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application.

[0198] In the embodiments provided by the present application, it should be understood that the disclosed apparatus / terminal device and method can be implemented in other ways. For example, the above-described apparatus / terminal device embodiments are only schematic. The division of the modules or units is only a logical function division, and there can be another division manner in actual implementation. For example, a plurality of units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the displayed or discussed mutual coupling or direct coupling or communication connection between each displayed or discussed module can be indirect coupling or communication connection through some interface, device or unit, and can be electrical, mechanical or in other forms.

[0199] The units described as separate components may or may not be physically separate, and the components displayed as units may or may not be physical units, that is, may be located in one place, or may also be distributed to multiple network units. Part or all of the units can be selected to achieve the purpose of the embodiment scheme according to actual needs.

[0200] The above embodiments are only used to illustrate the technical solutions of the present application, but not limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that the technical solutions recorded in the foregoing embodiments can still be modified, or some technical features can be replaced by equivalents; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application, and should be included in the protection scope of the present application.

Claims

1. A material proportioning intelligent optimization method fusing domain knowledge and machine learning, characterized in that, The method comprises: obtaining a plurality of training data sets of a fiber cement-based material, wherein each of the training data sets comprises at least training cement ratio, training sand-binder ratio and training fiber data of the fiber cement-based material; determining a target first regression coefficient of the fiber cement-based material based on the plurality of training data sets and a compressive strength formula of a cement-based material, the compressive strength formula being determined based on at least a first regression coefficient and a second regression coefficient of the cement-based material, the target first regression coefficient being used to calculate an actual compressive strength of the fiber cement-based material; constructing a first target prediction model of the fiber cement-based material according to the plurality of training data sets and the target first regression coefficient, the first target prediction model being used to predict an actual second regression coefficient of the fiber cement-based material, the actual second regression coefficient being used to calculate the actual compressive strength of the fiber cement-based material; constructing a second target prediction model of the fiber cement-based material according to the plurality of training data sets and the first target prediction model, the second target prediction model being used to predict the actual compressive strength of the fiber cement-based material; calculating a mix proportion of a material to be incorporated into the fiber cement-based material based on an objective function and a constraint condition of the fiber cement-based material, wherein at least part of the objective function is determined based on the actual compressive strength output by the second target prediction model and a target compressive strength; Wherein, each of the training data set further comprises: training water-binder ratio, training cement strength, training compressive strength, the compressive strength formula is Wherein, Representing compressive strength, Representing cement strength, W / B Representing water-binder ratio, Respectively represent the first regression coefficient and the second regression coefficient; the constructing of the first target prediction model of the fiber cement-based material according to the plurality of training data sets and the target first regression coefficient comprises: for each of the training data sets, inputting the training water-binder ratio, the training cement strength, the training compressive strength, and the target first regression coefficient of each of the training data sets into the compressive strength formula to obtain a theoretical second regression coefficient corresponding to each of the training data sets; establishing a first preliminary prediction model of the fiber cement-based material, the first preliminary prediction model being a random forest regressor; performing model training on the first preliminary prediction model based on the training water-binder ratio, the training cement strength, the training cement ratio, the training sand-binder ratio, and the training fiber data of the plurality of training data sets, and the theoretical second regression coefficient corresponding to each of the training data sets, to obtain the first target prediction model of the fiber cement-based material; the constructing of the second target prediction model of the fiber cement-based material according to the plurality of training data sets and the first target prediction model comprises: for each of the training data sets, inputting the training water-binder ratio, the training cement strength, the training cement ratio, the training sand-binder ratio, and the training fiber data of each of the training data sets into the first target prediction model to obtain an actual second regression coefficient corresponding to each of the training data sets; establishing a second preliminary prediction model of the fiber cement-based material, wherein the second preliminary prediction model is a multi-layer perception model, the multi-layer perception model has two hidden layers, each hidden layer contains 64 neurons, and an activation function of the multi-layer perception model is a ReLU function; model training is performed on the second preliminary prediction model according to each set of the training data set and the actual second regression coefficient corresponding to each set of the training data set, to obtain a second target prediction model of the fiber cement-based material.

2. The material proportioning intelligent optimization method fusing domain knowledge and machine learning according to claim 1, wherein, Each set of the training data set further includes a training component ratio and a training water reducer proportion; The model training performed on the second preliminary prediction model according to each set of the training data set and the actual second regression coefficient corresponding to each set of the training data set to obtain the second target prediction model of the fiber cement-based material includes: The second preliminary prediction model is determined as a second current prediction model of the fiber cement-based material; The training component ratio, the training water reducer proportion, the training water-binder ratio, the training sand-binder ratio, the training fiber data and the training cement strength of each set of the training data set are input into the second current prediction model to obtain a predicted compressive strength corresponding to each set of the training data set; The training water-binder ratio, the training cement strength, the target first regression coefficient and the actual second regression coefficient of each set of the training data set are input into the compressive strength formula to obtain a theoretical compressive strength corresponding to each set of the training data set; Model parameters of the second current prediction model are updated according to the predicted compressive strength and the theoretical compressive strength corresponding to each set of the training data set; The step of inputting the training component ratio, the training water reducer proportion, the training water-binder ratio, the training sand-binder ratio, the training fiber data and the training cement strength of each set of the training data set into the second current prediction model to obtain the predicted compressive strength corresponding to each set of the training data set is returned to be performed until the number of updates of the second current prediction model reaches a maximum iteration number; A second current prediction model corresponding to the maximum iteration number is determined as the second target prediction model of the fiber cement-based material.

3. The material proportioning intelligent optimization method fusing domain knowledge and machine learning according to claim 2, wherein, The updating of the model parameters of the second current prediction model according to the predicted compressive strength and the theoretical compressive strength corresponding to each set of the training data set includes: Based on the predicted compressive strength and the theoretical compressive strength corresponding to each set of the training data set, a physical loss value and a data loss value of each set of the training data set are respectively calculated; Based on the physical loss value and the data loss value of each set of the training data set, a total loss value of the multiple sets of training data sets is calculated; The model parameters of the second current prediction model are updated according to the total loss value.

4. The material proportioning intelligent optimization method fusing domain knowledge and machine learning according to claim 1, wherein, The determination of the target first regression coefficient of the fiber cement-based material based on the multiple sets of training data sets and a compressive strength formula of a cement-based material includes: determining a first initial value as a current first regression coefficient of the fiber cement-based material and a second initial value as a current second regression coefficient of the fiber cement-based material; inputting, for each of the training data sets, the training water-binder ratio, the training cement strength, the current first regression coefficient and the current second regression coefficient of each of the training data sets into the compressive strength formula to obtain a theoretical compressive strength of each of the training data sets; calculating a current residual sum of squares of the multiple training data sets based on the theoretical compressive strength and the training compressive strength of each of the training data sets; modifying the current first regression coefficient and the current second regression coefficient according to the current residual sum of squares, and returning to perform the step of inputting, for each of the training data sets, the training water-binder ratio, the training cement strength, the current first regression coefficient and the current second regression coefficient of each of the training data sets into the compressive strength formula to obtain a theoretical compressive strength of each of the training data sets until the current residual sum of squares reaches convergence or the number of modifications of the current first regression coefficient reaches a maximum iteration number; determining the current first regression coefficient corresponding to the current residual sum of squares reaching convergence or the number of modifications of the current first regression coefficient reaching the maximum iteration number as a target first regression coefficient of the fiber cement-based material.

5. The material proportioning intelligent optimization method of fusing domain knowledge and machine learning according to any one of claims 1-4, characterized in that, The target function includes a first target function aiming to minimize the compressive strength error, a second target function aiming to minimize the carbon emission, and a third target function aiming to minimize the material cost, the first target function being determined based on the actual compressive strength predicted by the second target prediction model and the target compressive strength; The calculation of the mix proportion of the incorporated material in the fiber cement-based material based on the target function and the constraint condition of the fiber cement-based material includes: determining a first function weight of the first target function, a second function weight of the second target function and a third function weight of the third target function, respectively; calculating the mix proportion of the incorporated material in the fiber cement-based material by using a non-dominated sorting genetic algorithm according to the first target function, the second target function, the third target function, the first function weight, the second function weight, the third function weight and the constraint condition, the mix proportion at least including a target cement strength, a target component ratio, a target water-reducing agent proportion, a target water-binder ratio, a target sand-binder ratio and target fiber data.

6. A material proportioning intelligent optimization device fusing domain knowledge and machine learning, characterized in that, The method includes: an acquisition module configured to acquire multiple training data sets of a fiber cement-based material, wherein each of the training data sets at least includes a training cement ratio, a training sand-binder ratio and training fiber data of the fiber cement-based material; a regression coefficient determination module, configured to determine a target first regression coefficient of the fiber cement-based material based on the multiple sets of training data sets and a compressive strength formula of the cement-based material, the compressive strength formula being determined based on at least a first regression coefficient and a second regression coefficient of the cement-based material, the target first regression coefficient being used to calculate an actual compressive strength of the fiber cement-based material; a first model construction module, configured to construct a first target prediction model of the fiber cement-based material according to the multiple sets of training data sets and the target first regression coefficient, the first target prediction model being used to predict an actual second regression coefficient of the fiber cement-based material, the actual second regression coefficient being used to calculate the actual compressive strength of the fiber cement-based material; a second model construction module, configured to construct a second target prediction model of the fiber cement-based material according to the multiple sets of training data sets and the first target prediction model, the second target prediction model being used to predict the actual compressive strength of the fiber cement-based material; a calculation module, configured to calculate a mix proportion of a material incorporated in the fiber cement-based material based on an objective function and a constraint condition of the fiber cement-based material, wherein the objective function is determined based on at least the actual compressive strength output by the second target prediction model and a target compressive strength; Wherein, each set of training data set further comprises: training water-cement ratio, training cement strength, training compressive strength, and the compressive strength formula is Wherein, represents the compressive strength, represents the cement strength, W / B represents the water-cement ratio, respectively represent the first regression coefficient and the second regression coefficient; The first model construction module is specifically configured to: for each set of training data sets, input a training water-binder ratio, a training cement strength, a training compressive strength and the target first regression coefficient of each set of training data sets into the compressive strength formula to obtain a theoretical second regression coefficient corresponding to each set of training data sets; establish a first preliminary prediction model of the fiber cement-based material, the first preliminary prediction model being a random forest regressor; perform model training on the first preliminary prediction model according to the training water-binder ratio, the training cement strength, the training cement ratio, the training sand-binder ratio and the training fiber data of the multiple sets of training data sets, and the theoretical second regression coefficient corresponding to each set of training data sets, to obtain the first target prediction model of the fiber cement-based material; The second model construction module includes: an input unit, configured to, for each set of training data sets, input a training water-binder ratio, a training cement strength, a training cement ratio, a training sand-binder ratio and training fiber data of each set of training data sets into the first target prediction model to obtain an actual second regression coefficient corresponding to each set of training data sets; a model establishment unit, configured to establish a second preliminary prediction model of the fiber cement-based material, wherein the second preliminary prediction model is a multi-layer perceptron model, the multi-layer perceptron model having two hidden layers, each hidden layer containing 64 neurons, and the activation function of the multi-layer perceptron model being a ReLU function; a model training unit, configured to perform model training on the second preliminary prediction model according to each set of training data sets and the actual second regression coefficient corresponding to each set of training data sets, to obtain the second target prediction model of the fiber cement-based material.

7. A terminal device comprising a processor, a memory, and a computer program stored in the memory and executable on the processor, characterized in that, The processor executes the computer program, so that the terminal device implements the method in any one of claims 1-5. The processor executes the computer program, so that the terminal device implements the method in any one of claims 1-5.

8. A computer program product, characterised in that, The computer program product, when running on the terminal device, causes the terminal device to perform the method of any one of claims 1-5.

Images