Three-graded concrete mix proportion optimization method, device and equipment and storage medium

Abstract

The invention provides a three-graded concrete mix proportion optimization method, device and equipment and a storage medium, and the method comprises the steps: obtaining multiple groups of sample data of three-graded concrete, and obtaining a training set based on all sample data; obtaining an associated physical equation set and an initial PINN model, determining a total loss function based on the associated physical equation set and the training set, and training the initial PINN model based on the total loss function to obtain a target PINN model; constructing a target function group based on the target PINN model, determining a target constraint condition, generating a Pareto solution set of the target function group through a non-dominated sorting genetic algorithm based on the target constraint condition and the target PINN model, determining the optimal mix proportion of the three-graded concrete based on the Pareto solution set, and outputting the optimal mix proportion. According to the technical scheme of the embodiment of the invention, the associated physical equation set is embedded into the target PINN model, effective training can be realized by using less sample data, and an unreasonable Pareto solution can be obtained by avoiding f, so that the required cost and the required time are reduced.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of mix proportion optimization of concrete, and in particular to a three-grade mix concrete mix proportion optimization method, device, equipment and storage medium. BACKGROUND

[0002] As the most widely used and largest amount of building material today, concrete plays an irreplaceable role in China's infrastructure construction. Concrete includes three-grade mix concrete, which is mainly used in major water conservancy projects such as dams and ship locks. With the rapid advancement of infrastructure construction, three-grade mix concrete is widely used in harsh environments. Under the synergistic effect of multiple environmental factors, three-grade mix concrete materials are prone to durability degradation. The durability degradation of concrete will lead to a significant decrease in the performance of the engineering structure, or the functional failure of the concrete structure will occur before the design service life is reached. Therefore, it is of great engineering application value to study the mix proportion of three-grade mix concrete.

[0003] In the prior art, the compressive strength is taken as a single optimization target, and the optimal mix proportion of three-grade mix concrete is predicted by a traditional machine learning model. A large amount of sample data is imported into the traditional machine learning model, and the traditional machine learning model obtains a prediction model by learning a large amount of sample data. Based on the prediction model, a plurality of Pareto solutions are obtained, and the optimal mix proportion is determined from all the Pareto solutions. However, there are Pareto solutions that violate the basic physical laws in the output Pareto solution set, which need to be screened by additional experiments, resulting in a long time and high cost required to determine the optimal mix proportion of three-grade mix concrete. SUMMARY

[0004] The present application aims to at least solve one of the technical problems existing in the prior art. To this end, the present application provides a three-grade mix concrete mix proportion optimization method, device, equipment and storage medium, which can reduce the required sample data and avoid unreasonable Pareto solutions by adding physical equations as constraint conditions, thereby reducing the required cost and time.

[0005] In a first aspect, an embodiment of the present application provides a three-grade mix concrete mix proportion optimization method, comprising: obtaining a plurality of groups of sample data of three-grade mix concrete, and obtaining a training set based on all the sample data; obtaining a set of associated physical equations and an initial PINN model, determining a total loss function based on the set of associated physical equations and the training set, and training the initial PINN model based on the total loss function to obtain a target PINN model, wherein the set of associated physical equations includes at least one physical equation related to three-grade mix concrete; construct a target function group based on the target PINN model, determine a target constraint condition, generate a Pareto solution set of the target function group based on the target constraint condition and the target PINN model by using a non-dominated sorting genetic algorithm, determine an optimal mix proportion of the three-grade concrete based on the Pareto solution set, and output the optimal mix proportion, wherein the Pareto solution set includes a plurality of Pareto solutions.

[0006] According to some embodiments of the present application, an associated physical equation set and an initial PINN model are obtained, a total loss function is determined based on the associated physical equation set and the training set, including: The associated physical equation set is obtained, wherein the associated physical equation set includes a mechanical constitutive equation, a heat conduction equation and a chloride ion diffusion equation, the mechanical constitutive equation, the heat conduction equation and the chloride ion diffusion equation are expressed in the form of partial differential equations, the expression of the mechanical constitutive equation is: The expression of the heat conduction equation is: The expression of the chloride ion diffusion equation is: , is stress, f is volume force per unit volume, T is temperature, p is density, c is specific heat capacity, k is thermal conductivity, Q is heat source term, t is time, D is chloride ion permeability coefficient, and C is chloride ion concentration; The initial PINN model is constructed, and the number of layers of the hidden layer and the number of neurons of each layer of the initial PINN model are determined according to the highest order of the mechanical constitutive equation, the heat conduction equation and the chloride ion diffusion equation; The total loss function is obtained based on the associated physical equation set and the initial PINN model.

[0007] According to some embodiments of the present application, the total loss function is obtained based on the associated physical equation set and the initial PINN model, including: Residuals of the mechanical constitutive equation, the heat conduction equation and the chloride ion diffusion equation are respectively determined based on the associated physical equation set and the initial PINN model, to obtain mechanical residual, heat conduction residual and chloride ion diffusion residual; Mechanical loss term, heat conduction loss term and chloride ion diffusion loss term are respectively obtained based on a mean square error formula, the mechanical residual, the heat conduction residual and the chloride ion diffusion residual, wherein the expression of the mean square error formula is: The expression of the mechanical loss term is: The expression of the heat conduction loss term is: The expression of the chloride ion diffusion loss term is: n is the number of all sample data in the training set, a true value of an i th sample data in the training set, a predicted value of the i th sample data in the training set, the mechanical loss term, the heat conduction loss term, the chloride ion diffusion loss term, a number of region sampling points in a definition domain, and i and n are positive integers; determining an initial condition loss term and a boundary condition loss term, wherein an expression of the initial condition loss term is: an expression of the boundary condition loss term is: wherein, the initial condition loss term, the boundary condition loss term, a number of region sampling points of the initial condition, a number of region sampling points of the boundary condition, an i th sample data in all the sample data, obtaining the total loss function based on the mechanical loss term, the heat conduction loss term, the chloride ion diffusion loss term, the boundary condition loss term and the initial condition loss term, wherein an expression of the total loss function is: , a weight of the mechanical loss term, a weight of the heat conduction loss term, a weight of the chloride ion diffusion loss term, a weight of the initial condition loss term, a weight of the boundary condition loss term.

[0008] According to some embodiments of the present application, a plurality of sample data of a three-grade concrete is obtained, and a training set is obtained based on all the sample data, comprising: obtaining a mix proportion material and a performance index of the three-grade concrete, and obtaining all the sample data based on the mix proportion material and the performance index, wherein the performance index comprises a concrete compressive strength, a chloride ion permeability coefficient and an adiabatic temperature rise value; performing data cleaning on all the sample data, and performing a normalization operation on all the sample data based on a normalization formula to obtain a total data set, wherein the normalization formula is: , a normalized value, an i th sample data in all the sample data, a j th sample data in all the sample data, a minimum value in all the sample data, is the maximum value in all the sample data, e is the number of all the sample data, e and j are positive integers; The total data set is divided to obtain the training set and the test set.

[0009] According to some embodiments of the application, after the initial PINN model is trained based on the total loss function to obtain a target PINN model, the method further comprises: obtaining the test set, and obtaining predicted values of all the sample data in the test set based on the test set and the target PINN model; obtaining true values of all the sample data in the test set, and respectively obtaining a root mean square error (RMSE), a determination coefficient (R2) and a mean absolute error (MAE) based on all the predicted values in the test set, all the true values in the test set, a root mean square error formula, a determination coefficient formula and a mean absolute error formula, wherein an expression of the root mean square error formula is: an expression of the determination coefficient is: and an expression of the mean absolute error is: RMSE is the root mean square error, z is the number of the sample data in the test set, is the true value of the mth sample data in the test set, is the predicted value of the mth sample data in the test set, is the determination coefficient, is the average value of the true values of the sample data, MAE is the mean absolute error, z and m are positive integers, and m is less than or equal to z; When the determination coefficient is greater than a first preset threshold, the root mean square error is less than a second preset threshold, and the mean absolute error is less than a third preset threshold, the fitting accuracy and the prediction accuracy of the target PINN model meet the requirements.

[0010] According to some embodiments of the application, a target function group is constructed based on the target PINN model, and a target constraint condition is determined, comprising: The target function group is constructed based on the target PINN model, wherein the target function group comprises a compression resistance target function, a chloride ion permeability coefficient target function, an adiabatic temperature rise value target function and an economic cost target function, an expression of the compression resistance target function is: an expression of the chloride ion permeability coefficient is: an expression of the adiabatic temperature rise value target function is and an expression of the economic cost target function is: , is a regression function of the compressive strength of the three-grade concrete, is a regression function of the chloride ion permeability coefficient of the three-grade concrete, is a regression function of the adiabatic temperature rise of the three-grade concrete, is a regression function of the economic cost of the three-grade concrete, X is a parameter variable of the mix material, the parameter variable of the mix material includes a water-binder ratio, a cement dosage, a sand dosage, a gravel dosage, a fly ash dosage, a slag powder dosage, and a water reducing agent dosage, is the water-binder ratio, is the cement dosage, is the sand dosage, is the gravel dosage, is the fly ash dosage, is the slag powder dosage, is the water reducing agent dosage, is a unit mass cost price of the parameter variable of the mix material, is the i-th parameter variable of the mix material; an initial constraint condition is constructed, wherein an expression of the initial constraint condition is: , is the i-th parameter variable of the mix material, is a lower limit of the normative design of the i-th parameter variable of the mix material, is an upper limit of the normative design of the i-th parameter variable of the mix material; the target constraint condition is obtained based on the initial constraint condition, wherein an expression of the target constraint condition is: .

[0011] According to some embodiments of the present application, a Pareto solution set of the target function group is generated based on the target constraint condition and the target PINN model by a non-dominated sorting genetic algorithm, an optimal mix proportion of three-grade concrete is determined based on the Pareto solution set, and the optimal mix proportion is output, including: a multi-factor optimization of the target PINN model is performed by the non-dominated sorting genetic algorithm, and an iteration termination condition is set, wherein the iteration termination condition includes a size of an initial population, a maximum iteration number, and a convergence standard Pareto change rate; the initial population is generated based on the iteration termination condition, wherein the initial population includes a plurality of individuals, each individual is used to represent a group of parameter combinations of the mix proportion of the three-grade concrete, and the individual satisfies the target constraint condition; Based on the target function set, fitness evaluation and non-dominated sorting are performed on all the individuals of the initial population, to obtain fitness evaluation results and non-dominated sorting results, and the Pareto solution set is determined through the fitness evaluation results and the non-dominated sorting results; The optimal mix proportion is determined based on all the Pareto solutions of the Pareto solution set, and the optimal mix proportion is output.

[0012] In a second aspect, an embodiment of the present application provides a three-grade concrete mix proportion optimization device, which comprises at least one control processor and a memory connected in communication with the at least one control processor; the memory stores instructions executable by the at least one control processor, and the instructions are executed by the at least one control processor to enable the at least one control processor to perform the three-grade concrete mix proportion optimization method as described in the first aspect.

[0013] In a third aspect, an embodiment of the present application provides an electronic device comprising the three-grade concrete mix proportion optimization device as described in the second aspect.

[0014] In a fourth aspect, an embodiment of the present application provides a computer readable storage medium storing computer executable instructions for performing the three-grade concrete mix proportion optimization method as described in the first aspect.

[0015] According to the three-grade concrete mixing proportion optimization method provided in the embodiments of the present application, the following beneficial effects are achieved: a plurality of sample data sets of three-grade concrete are obtained, a training set is obtained based on all the sample data sets, a correlation physical equation set and an initial PINN model are obtained, a total loss function is determined based on the correlation physical equation set and the training set, the initial PINN model is trained based on the total loss function to obtain a target PINN model, wherein the correlation physical equation set comprises at least one physical equation related to three-grade concrete; a target function set is constructed based on the target PINN model, a target constraint condition is determined, a Pareto solution set of the target function set is generated by a non-dominated sorting genetic algorithm based on the target constraint condition and the target PINN model, an optimal mixing proportion of three-grade concrete is determined based on the Pareto solution set, and the optimal mixing proportion is output, wherein the Pareto solution set comprises a plurality of Pareto solutions. According to the technical solution of the embodiments of the present application, the total loss function based on the correlation physical equation set is fused in the initial PINN model, the total loss function adds a physical constraint condition to the target PINN model, the interval of the Pareto solution output by the target PINN model is accurately limited, the fitting accuracy of the target function set is effectively improved, the dependence of the target PINN model on the sample data is reduced, the Pareto solutions output all conform to basic physical laws, the prediction behavior of the target PINN model in the sparse area of the sample data is enhanced, the generalization ability of the target PINN model to unknown materials and environmental conditions is enhanced, and therefore, when the optimal mixing proportion of three-grade concrete is determined by the target PINN model, the required cost is reduced and the required time is shortened. BRIEF DESCRIPTION OF DRAWINGS

[0016] Figure 1 is a flowchart of a three-grade concrete mixing proportion optimization method provided in an embodiment of the present application; Figure 2 is a structural diagram of a three-grade concrete mixing proportion optimization device provided in another embodiment of the present application. DETAILED DESCRIPTION

[0017] Embodiments of the present application are described in detail below, examples of which are shown in the accompanying drawings, wherein the same or similar reference signs represent the same or similar elements or elements having the same or similar functions throughout. The embodiments described below by reference to the accompanying drawings are exemplary and are only used to explain the present application, and cannot be understood as a limitation of the present application.

[0018] In the description of the present application, it should be understood that the orientation description, such as the orientation or position relationship indicated by the upper, lower, front, rear, left, right and the like, is based on the orientation or position relationship shown in the drawings, and is only for the convenience of describing the present application and simplifying the description, and cannot be understood as indicating or implying that the device or element indicated must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present application.

[0019] In the description of the present application, one or more is understood as one or more, more than two is understood as more than two, greater than, less than, more than, etc. are understood as not including the number, above, below, etc. are understood as including the number. If the first, second is described, it is only for the purpose of distinguishing technical features, and cannot be understood as indicating or implying relative importance or implicitly indicating the number of indicated technical features or the sequence of indicated technical features.

[0020] In the description of the present application, unless otherwise explicitly limited, the words such as setting, installing, connecting and the like should be understood in a broad sense, and those skilled in the art can reasonably determine the specific meaning of the above words in the present application in combination with the specific content of the technical solution.

[0021] The three-grade concrete mixing ratio optimization method according to the embodiment of the present application has at least the following beneficial effects: a plurality of groups of sample data of three-grade concrete are obtained, a training set is obtained based on all the sample data; an associated physical equation set and an initial PINN model are obtained, a total loss function is determined based on the associated physical equation set and the training set, the initial PINN model is trained based on the total loss function to obtain a target PINN model, wherein the associated physical equation set includes at least one physical equation related to three-grade concrete; a target function set is constructed based on the target PINN model, a target constraint condition is determined, a Pareto solution set of the target function set is generated based on the target constraint condition and the target PINN model by using a non-dominated sorting genetic algorithm, an optimal mixing ratio of three-grade concrete is determined based on the Pareto solution set, and the optimal mixing ratio is output, wherein the Pareto solution set includes a plurality of Pareto solutions. According to the technical scheme of the embodiment of the present application, the total loss function based on the associated physical equation set is fused in the initial PINN model, the total loss function adds a physical constraint condition to the target PINN model, can accurately limit the interval of the Pareto solution output by the target PINN model, thereby effectively improving the fitting accuracy of the target function set, reducing the dependence of the target PINN model on the sample data, and the output Pareto solution conforms to the basic physical law, and can enhance the prediction behavior of the target PINN model in the sparse area of the sample data, enhance the generalization ability of the target PINN model to unknown materials and environmental conditions, so that when the optimal mixing ratio of three-grade concrete is determined by using the target PINN model, the required cost is reduced and the required time is shortened.

[0022] The following drawings further illustrate the technical solutions of the embodiments of the present application.

[0023] Referring to Figure 1 , Figure 1 A flowchart of a three-grade mixed concrete mix proportion optimization method provided by an embodiment of the present application is shown in FIG. 1. The three-grade mixed concrete mix proportion optimization method includes but is not limited to the following steps: S10, obtaining a plurality of groups of sample data of three-grade mixed concrete, and obtaining a training set based on all the sample data; S20, obtaining a set of associated physical equations and an initial PINN model, determining a total loss function based on the set of associated physical equations and the training set, training the initial PINN model based on the total loss function to obtain a target PINN model, wherein the set of associated physical equations includes at least one physical equation related to three-grade mixed concrete; S30, constructing a set of objective functions based on the target PINN model, determining a target constraint condition, generating a Pareto solution set of the set of objective functions based on the target constraint condition and the target PINN model by using a non-dominated sorting genetic algorithm, determining an optimal mix proportion of three-grade mixed concrete based on the Pareto solution set, and outputting the optimal mix proportion, wherein the Pareto solution set includes a plurality of Pareto solutions.

[0024] It should be noted that the mix proportion of the three-grade mixed concrete is adjusted manually to obtain a plurality of three-grade mixed concrete with different mix proportions, and all sample data is obtained based on all the three-grade mixed concrete with different mix proportions. The total data set is obtained by data cleaning and normalization processing of all the sample data, and the training set is obtained by dividing the total data set.

[0025] It should be noted that the set of associated physical equations is a physical equation related to three-grade mixed concrete, such as a physical equation related to the compressive strength, durability and compressive resistance of concrete. The total loss function is determined based on the set of associated physical equations and the training set, and the total loss function is used to train the initial PINN model to obtain the target PINN model, so that the target function output by the obtained target PINN model conforms to the physical law.

[0026] It should be noted that after obtaining the total loss function, the initial PINN model is trained using a neural network Adam and L-BFGS optimizer hybrid strategy to obtain the target PINN model. After obtaining the target PINN model, the hyperparameters of the target PINN model are optimized using the Bayesian optimization theory.

[0027] It should be noted that, since the existing second-generation non-dominated sorting genetic algorithm can only be applied to the case where the number of objective functions is less than three, when more than 3 or more objectives are solved, the problems of degeneration of Pareto front and uneven distribution of Pareto solution are prone to occur, therefore, the second-generation non-dominated sorting genetic algorithm is difficult to be applied to multi-objective tasks when solving multi-objective optimization, therefore, in the present application, the third-generation non-dominated sorting genetic algorithm is adopted, so as to realize multi-factor optimization of multiple objective functions in the objective function group.

[0028] It should be noted that, based on the objective function group and the target constraint condition, the multi-objective optimization is carried out based on the non-dominated sorting genetic algorithm, and the global optimization is carried out, so as to obtain a Pareto solution set, the Pareto solution set includes multiple Pareto solutions, and the Pareto solution is the mix proportion of the three-grade concrete meeting the requirements of the objective function group and the target constraint condition; that is, under the condition of meeting the requirements of the objective function group and the target constraint condition, the mix proportion scheme of the three-grade concrete is calculated, the priority is determined based on the multi-dimensional demand for the three-grade concrete, the weight of all demands is determined based on the priority sorting, the obtained Pareto solution is scored based on the priority and the weight, and all Pareto solutions are sorted based on the scoring results, and the optimal mix proportion is determined based on the sorting results.

[0029] It should be noted that in the existing method, the concrete mix proportion optimization model relies on artificial periodic retraining of the model, and cannot respond to changes in data distribution in a timely manner, and the model lacks autonomous updating capability. Therefore, the three-grade concrete mix proportion optimization method of the present application can realize autonomous updating, can realize real-time dynamic collection of the verification results of the sample data and the engineering data of the three-grade concrete, and continuously updates and iterates the target PINN model through active learning and online learning mechanism, realizes the closed-loop continuous optimization design of "data-model-decision", solves the problem of static design of traditional mix proportion, realizes the autonomous optimization and dynamic real-time of the mix proportion design of the three-grade concrete. Compared with the existing method, the present application solves the deficiencies of the prior art in real-time, safety and resource efficiency by constructing an autonomous updating system of the target PINN model, and the obtained target PINN model has physical meaning, high prediction accuracy and generalization ability, so that the three-grade concrete mix proportion optimization method of the present application has higher stability and accuracy.

[0030] In the prior art, a traditional machine learning model is used as a concrete mix proportion optimization model to predict the optimal mix proportion of the three-grade concrete. However, the existing concrete mix proportion optimization model is purely data-driven, that is, the concrete mix proportion optimization model completely relies on the imported sample data, and a large amount of sample data is required to cover all possible mix combinations and environmental conditions. Otherwise, the concrete mix proportion optimization model is prone to underfitting or overfitting, or the prediction outside the sample data is unreliable, or the concrete mix proportion optimization model learns the noise or pseudo-correlation in the sample data, resulting in the existence of Pareto solutions that violate the basic physical laws in the output Pareto solution set, which requires additional experimental screening. The three-grade concrete mix proportion material design has multiple components (such as cement, aggregate, and admixture) and complex physical and chemical processes (such as hydration, diffusion, and strength development process), resulting in high cost and long cycle for determining the optimal mix proportion of the three-grade concrete.

[0031] In the present application, the total loss function based on the associated physical equation set is fused in the initial PINN model, which adds physical constraint conditions to the target PINN model, accurately limits the interval of the Pareto solution output by the target PINN model, effectively improves the fitting accuracy of the target function group, reduces the dependence of the target PINN model on sample data, and the output Pareto solution conforms to the basic physical laws. In addition, the prediction behavior of the target PINN model in the sparse area of the sample data is enhanced, and the generalization ability of the target PINN model to unknown materials and environmental conditions is enhanced, so that when the optimal mix proportion of the three-grade concrete is determined by the target PINN model, the required cost is reduced and the required time is shortened.

[0032] In addition, in an embodiment, in Figure 1 In step S20 shown in the figure, the associated physical equation set and the initial PINN model are obtained, and the total loss function is determined based on the associated physical equation set and the training set, including but not limited to the following steps: S21, the associated physical equation set is obtained, wherein the associated physical equation set includes the mechanical constitutive equation, the heat conduction equation and the chloride ion diffusion equation, the mechanical constitutive equation, the heat conduction equation and the chloride ion diffusion equation are expressed in the form of partial differential equation, the expression of the mechanical constitutive equation is: , the expression of the heat conduction equation is: , and the expression of the chloride ion diffusion equation is: , where σ is the stress, f is the body force per unit volume, T is the temperature, ρ is the density, c is the specific heat capacity, k is the thermal conductivity, Q is the heat source term, t is the time, D is the chloride ion permeability coefficient, and C is the chloride ion concentration. S22, constructing an initial PINN model, determining the number of layers of the hidden layer and the number of neurons of each layer of the initial PINN model according to the highest order of the mechanical constitutive equation, the heat conduction equation and the chloride ion diffusion equation; S23, obtaining a total loss function based on the associated physical equation set and the initial PINN model.

[0033] It should be noted that the mechanical constitutive physical equation determines the deformation of the concrete caused by stress, the thermal conduction physical equation determines the performance difference of the concrete caused by temperature change, and the diffusion equation physical equation determines the corrosion of the concrete caused by chloride ion diffusion. The control equation of the deformation of the concrete caused by stress is determined by using the linear elastic constitutive equation; the control equation of the performance difference of the concrete caused by temperature change is determined by using the heat conduction equation; and the control equation of the corrosion process of the concrete caused by chloride ion is determined by using the second Fick's law to describe the diffusion process of the chloride ion in the concrete. The associated physical equation set includes the mechanical constitutive equation, the heat conduction equation and the chloride ion diffusion equation.

[0034] It should be noted that the linear elastic constitutive equation is used as the control equation of the material mechanical behavior, and the mechanical constitutive equation is usually used to describe the stress-strain relationship of the material under external load, and reflects the mechanical properties such as elasticity, plasticity or viscosity of the material. The mechanical response of the concrete material can be described by the linear elastic constitutive equation, and the expression of the linear elastic constitutive equation is: , wherein, is the stress, is the strain, is the elastic tensor; the mechanical constitutive equation is determined based on the linear elastic constitutive equation, the mechanical constitutive equation is expressed in the form of partial differential equation, and the expression of the mechanical constitutive equation is: .

[0035] It should be noted that the Fourier heat conduction equation is used as the control equation of the heat transfer process, and the heat conduction equation is used to simulate the change law of the temperature field inside the object with time and space, and the transient heat conduction process can be represented by the heat conduction equation expressed in the form of partial differential equation, and the expression of the heat conduction equation is: , x is the position.

[0036] It should be noted that the diffusion-reaction equation of chloride ion erosion of concrete is used as the control equation of the chloride ion erosion process of concrete, and the diffusion-reaction equation of chloride ion erosion of concrete is usually used to simulate the damage and corrosion of concrete in the chloride ion erosion environment, and the chloride ion penetration process of concrete can be described by the second Fick's law, and the expression of the chloride ion diffusion equation expressed in the form of partial differential equation is: , x is the position.

[0037] It should be noted that embedding the physical equations such as the mechanical constitutive equation, the heat conduction equation and the chloride ion diffusion equation as physical constraints into the total loss function can reduce the dependence of the target PINN model on sample data, enhance the physical meaning and interpretability of the target PINN model, make the target function output by the target PINN model conform to the basic physical law, and thus enhance the prediction accuracy of the target PINN model in the performance indicators of compressive strength, durability and crack resistance.

[0038] In addition, in an embodiment, in step S23, the following steps are included but not limited to: S231, based on the associated physical equation set and the initial PINN model, the residual errors of the mechanical constitutive equation, the heat conduction equation and the chloride ion diffusion equation are determined respectively to obtain the mechanical residual error, the heat conduction residual error and the chloride ion diffusion residual error; S232, based on the mean square error formula, the mechanical residual error, the heat conduction residual error and the chloride ion diffusion residual error, the mechanical loss term, the heat conduction loss term and the chloride ion diffusion loss term are obtained respectively, wherein the expression of the mean square error formula is: , the expression of the mechanical loss term is: , the expression of the heat conduction loss term is: , the expression of the chloride ion diffusion loss term is: n is the number of all sample data in the training set, is the true value of the i th sample data in the training set, is the predicted value of the i th sample data in the training set, is the mechanical loss term, is the heat conduction loss term, is the chloride ion diffusion loss term, n and i are positive integers; S233, the initial condition loss term and the boundary condition loss term are determined, wherein the expression of the initial condition loss term is: , the expression of the boundary condition loss term is: , wherein is the initial condition loss term, is the boundary condition loss term, is the number of region selected points of the initial condition, is the number of region selected points of the boundary condition, and is the i th sample data in the all sample data; S234, based on the mechanical loss term, the heat conduction loss term, the chloride ion diffusion loss term, the boundary condition loss term and the initial condition loss term, the total loss function is obtained, wherein the expression of the total loss function is: , is the weight of the mechanical loss term, The weight of the heat conduction loss term, The weight of the chloride ion diffusion loss term. The weights of the initial conditional loss term, The weights of the boundary condition loss term.

[0039] It should be noted that, , , , and All values ​​are set to 1 to improve the training accuracy and speed of the initial PINN model.

[0040] It should be noted that, through the automatic differentiation function of deep learning frameworks such as PyTorch or TensorFlow, the linear elastic constitutive equation, Fourier heat conduction formula, and Fick's second law formula are converted into mechanical constitutive equation, heat conduction equation, and chloride ion diffusion equation, respectively. The mechanical constitutive equation, heat conduction equation, and chloride ion diffusion equation are expressed in the form of partial differential equations. Substituting the mechanical constitutive equation, heat conduction equation, and chloride ion diffusion equation into the output of the initial PINN model, the residuals are calculated respectively. The mean square error of all the obtained residuals is used as the loss term, thus obtaining the mechanical loss term, heat conduction loss term, and chloride ion diffusion loss term. The obtained mechanical loss term, heat conduction loss term, and chloride ion diffusion loss term are combined with the boundary conditions and initial conditions to obtain the total loss function.

[0041] Additionally, in one embodiment, in Figure 1 Step S10 shown includes, but is not limited to, the following steps: S11, obtain the mix proportion materials and performance indicators of the three-grade concrete, and obtain all sample data based on the mix proportion materials and performance indicators. Among them, the performance indicators include concrete compressive strength, chloride ion permeability coefficient and adiabatic temperature rise value. S12, perform data cleaning on all sample data, and normalize all sample data according to the normalization formula to obtain the total dataset. The normalization formula is: , Normalized value For the i-th sample data in all the sample data, For the j-th sample data in all sample data, The minimum value among all sample data. Let j be the maximum value among all sample data, e be the number of all sample data, and both e and j are positive integers; S13, the total dataset is divided into training set and test set.

[0042] It should be noted that the performance indicators include the compressive strength of concrete, the chloride ion permeability coefficient and the adiabatic temperature rise value, wherein the compressive strength of concrete is used to characterize the strength of the three-grade concrete, the chloride ion permeability coefficient is used to characterize the durability of the three-grade concrete, and the adiabatic temperature rise value is used to characterize the crack resistance of the three-grade concrete.

[0043] It should be noted that the sample data is obtained based on the performance indicators, which can avoid the prior art in which only the compressive performance of concrete is used as a single and unique optimization target, so that the optimal mix ratio obtained can meet the performance requirements of the compressive strength of concrete, the chloride ion permeability coefficient and the adiabatic temperature rise value, thereby ensuring the quality and service life of the three-grade concrete in various harsh environments.

[0044] It should be noted that the mix ratio materials of the three-grade concrete include water-binder ratio, cement, sand, stone, water reducing agent, fly ash and slag powder, wherein the stone includes 5-20 mm gravel, 20-40 mm gravel and 40-80 mm gravel.

[0045] It should be noted that the sample data comes from detection equipment in the laboratory and sensors applied to the engineering site with three-grade concrete, and the application can obtain the material performance of the three-grade concrete and the environmental data of the engineering site in real time, wherein the sensors of the engineering site include temperature and humidity sensors, strain gauges and other equipment; the detection equipment in the laboratory and the sensors of the engineering site can automatically collect the mix ratio and environmental data of the three-grade concrete; the environmental data includes temperature, humidity, chloride ion concentration and related data of erosion medium. Since the sample data comes from various sources, it is necessary to clean and normalize the sample data to realize format standardization, unify the structured format, and uniformly convert the sample data from different sources into an analyzable structured table, thereby providing high-quality input for subsequent training of the initial PINN model, avoiding the initial PINN model from learning incorrect sample data, and causing the prediction accuracy of the target PINN model obtained by training to be low.

[0046] It should be noted that all sample data is cleaned, wherein the data cleaning includes cleaning data outliers and / or missing values in the sample data. By normalizing the sample data cleaned to the interval [-1, 1], the purpose is to unify the specifications of various sample data, eliminate the dimensional differences, and avoid the data in the sample data being too large or too small, which affects the prediction accuracy and convergence performance of the subsequent target PINN model.

[0047] It should be noted that the source of the sample data is comprehensively used by means of sensor monitoring, exposure test, prototype observation, field detection and other means, and a large amount of data under multiple working conditions and long time sequence is accumulated to provide a reliable data basis for model training; secondly, the feature extraction means is used to screen out the key parameters, and the characteristics of various sample data are fully considered, the multi-source data fusion is constructed, the preprocessing process of the multi-source data is carried out, the data cleaning link, the abnormal and repeated information processing, the missing value is filled by using the sparse data, and the integrity and reliability of the sample data are ensured, the data standardization processing is carried out, the specifications of various data are unified, and the dimension difference is eliminated; the data correlation analysis is carried out, and the internal relation between the data is deeply mined; finally, the data set is divided according to the scientific principle, and the adaptive data resources are provided for model training and verification, and the total data set is divided into 80% as the training set and 20% as the test set.

[0048] It should be noted that in the present application, the form of the sample data used is shown in Table 1 as follows:

[0049] Table 1: Sample data of three-level concrete mix proportion In addition, in an embodiment, after the step S20 shown in the figure, the following steps are included but not limited to: Figure 1 The step S20 shown in the figure, the following steps are included but not limited to: S201, obtaining a test set, and obtaining predicted values of all sample data in the test set based on the test set and a target PINN model; S202, obtaining true values of all sample data in the test set, and respectively obtaining a root mean square error, a determination coefficient and an average absolute error based on all predicted values in the test set, all true values in the test set, a root mean square error formula, a determination coefficient formula and an average absolute error formula, wherein the expression of the root mean square error formula is: The expression of the determination coefficient is: The expression of the average absolute error is: RMSE is the root mean square error, z is the number of sample data in the test set, is a true value of the mth sample data in the test set, is a predicted value of the mth sample data in the test set, is the determination coefficient, is the average value of the true value of the sample data, MAE is the average absolute error, z and m are positive integers, and m is less than or equal to z; S203, when the determination coefficient is greater than a first preset threshold, the root mean square error is less than a second preset threshold, and the average absolute error is less than a third preset threshold, the fitting precision and the prediction accuracy of the target PINN model meet the requirements.

[0050] It should be noted that the target PINN model is tested using the root mean square error formula, coefficient of determination formula, mean absolute error formula, and a test set to determine whether the fitting accuracy and prediction accuracy of the obtained target PINN model meet the requirements. If the root mean square error, coefficient of determination, or mean absolute error does not meet the fitting accuracy requirements, it is necessary to check whether there are abnormalities in the sample data or problems with the dispersion of the sample data.

[0051] Additionally, in one embodiment, in Figure 1 In step S30 shown, a set of objective functions is constructed based on the target PINN model, and the objective constraints are determined, including but not limited to the following steps: S31. Construct a set of objective functions based on the target PINN model. This set includes the compressive strength objective function, the chloride ion permeability coefficient objective function, the adiabatic temperature rise objective function, and the economic cost objective function. The expression for the compressive strength objective function is as follows: The expression for the chloride ion permeability coefficient is: The expression for the objective function of adiabatic temperature rise is: The expression for the economic cost objective function is: , The regression function for the compressive strength of three-graded concrete is given. This is a regression function for the chloride ion permeability coefficient of three-grade concrete. The regression function for the adiabatic temperature rise of three-grade concrete. Let X be the regression function for the economic cost of three-grade concrete, where X represents the parameter variables of the mix proportion materials, including water-cement ratio, cement content, sand content, crushed stone content, fly ash content, slag powder content, and water-reducing agent content. This refers to the water-to-glue ratio. This refers to the amount of cement used. The amount of sand used. This refers to the amount of crushed stone used. This refers to the amount of fly ash used. This refers to the amount of slag powder used. For the dosage of water-reducing agent, The unit mass cost price of the materials used in the mix proportion. Let i be the parameter variable of the i-th mix proportion material; S32, Construct initial constraints, where the expressions for the initial constraints are: , Let i be the parameter variable of the i-th mix proportion material. The lower limit of the parameter variables for the i-th mix proportion material is the specification design limit. The upper limit of the parameter variables for the i-th mix proportion material in the specification design; S33, obtaining a target constraint condition based on the initial constraint condition, wherein an expression of the target constraint condition is: .

[0052] It should be noted that, in the process of concrete mix proportioning, in order to meet the requirements of relevant specifications, according to the actual engineering and the modified specification requirements, the value range of concrete materials and the mix proportion relationship between each mix proportion material are taken as constraint conditions to establish the constraint range of the parameters of concrete mix proportioning.

[0053] It should be noted that, by Pytorch programming code, a target function group is obtained based on the target PINN model, and the target PINN model and the economic cost of the third-grade mixed concrete are taken as the target function of the third-grade mixed concrete. By the third generation of non-dominated sorting genetic algorithm, according to the set iteration termination condition, a multi-factor global search of the strength, durability and crack resistance of the third-grade mixed concrete is carried out, so as to generate a Pareto solution set of the target function group.

[0054] It should be noted that the concrete performance to be balanced and optimized includes: concrete compressive strength, chloride ion permeability coefficient, adiabatic temperature rise value and economic cost; according to the constraint range of the parameter variables of each mix proportion material set by the actual engineering demand and the standard specification.

[0055] It should be noted that by embedding the mechanical constitutive equation, the heat conduction equation and the chloride ion diffusion equation into the total loss function as physical constraints, combining with sample data, a data-physical double-driven target function group is constructed, and by using adaptive weight optimization and mixed training strategy, the collaborative cross-scale precise prediction of the compressive strength, durability and crack resistance of the third-grade mixed concrete is realized, which can significantly reduce the dependence on sample data, enhance the physical meaning and interpretability of the target PINN model, and provide high-precision target function group for determining the optimal mix proportion.

[0056] In addition, in an embodiment, in Figure 1 As shown in step S30, based on the target constraint condition and the target PINN model, a Pareto solution set of the target function group is generated by the non-dominated sorting genetic algorithm, and the optimal mix proportion of the third-grade mixed concrete is determined based on the Pareto solution set, and the optimal mix proportion is output, including but not limited to the following steps: S34, the target PINN model is optimized by the non-dominated sorting genetic algorithm, and the iteration termination condition is set, wherein the iteration termination condition includes the size of the initial population, the maximum number of iterations and the convergence standard Pareto change rate; S35, generating an initial population based on the iteration termination condition, wherein the initial population includes a plurality of individuals, each individual is used to represent a group of parameter combinations of the mix proportion of the third-grade mixed concrete, and the individual satisfies the target constraint condition; S36, perform fitness evaluation and non-dominated sorting on all individuals of the initial population based on the objective function set to obtain fitness evaluation results and non-dominated sorting results, and determine a Pareto solution set through the fitness evaluation results and the non-dominated sorting results; S37, determine an optimal mix proportion based on all Pareto solutions in the Pareto solution set, and output the optimal mix proportion.

[0057] It should be noted that the target PINN model is optimized by the third generation non-dominated sorting genetic algorithm, and the Pareto solution set of the objective function set is output according to the iteration termination condition set in the third generation non-dominated sorting genetic algorithm model. The Pareto optimal solution is determined based on all Pareto solutions in the Pareto solution set, that is, the optimal mix proportion of the three-grade concrete.

[0058] It should be noted that the third generation non-dominated sorting genetic algorithm specifically includes the following steps: randomly generating an initial population Y with a size of N, the initial population Y including multiple individuals, the individuals being used to represent a group of parameter combinations of the mix proportion of the three-grade concrete, and the individuals satisfying the constraint condition limit and satisfying the requirements of the objective function set; performing fitness evaluation and non-dominated sorting on all individuals in the initial population Y according to the objective function set; obtaining multiple non-dominated levels, each non-dominated level including at least one individual; based on any one non-dominated level, calculating the crowding distance of the individuals in the non-dominated level; selecting parent individuals based on the non-dominated sorting and the crowding distance, generating offspring using crossover operations such as single-point crossover or uniform crossover, generating new individuals by exchanging the genes of the parent individuals, and randomly changing the genes of some individuals to improve the randomness of the population. Fitness evaluation results are obtained by performing fitness evaluation, which can help the algorithm select the most representative solution and ensure that the diversity of the population is maintained during the optimization process, avoiding falling into local optimum.

[0059] It should be noted that the Pareto solution set of the objective function set for optimizing the mix proportion of the three-grade concrete is output by the algorithm iteration based on the non-dominated sorting and the crowding distance and the iteration termination condition, and the Pareto optimal solution is determined based on all Pareto solutions in the Pareto solution set, that is, the optimal mix proportion of the three-grade concrete.

[0060] It should be noted that the multi-factor optimization design of the third generation non-dominated sorting genetic algorithm is called to generate different mix proportion schemes of the mix proportion materials of the three-grade concrete that meet the actual engineering requirements, that is, to generate the Pareto solution set, under the premise of ensuring the strength, durability, crack resistance and economy of the three-grade concrete. According to the weight scoring mechanism, all mix proportion schemes in the Pareto solution set are comprehensively scored to determine the optimal mix proportion.

[0061] It should be noted that according to the multi-dimensional requirements of the strength, durability, crack resistance and economic cost of the three-grade concrete, the priorities are set, the weights of different requirements are set based on the priorities, and the strength, durability, crack resistance and economic cost of the three-grade concrete are weighted and scored, so as to select the optimal mix proportion required by the actual engineering.

[0062] In order to understand the technical scheme of the present application, the following specific embodiments are provided: S401, obtaining the mix proportion material and performance index of the three-grade concrete, obtaining all sample data based on the mix proportion material and performance index, performing data cleaning on all sample data, and performing normalization operation on all sample data based on a normalization formula to obtain a total data set, and dividing the total data set to obtain a training set and a test set, wherein the performance index includes concrete compressive strength, chloride ion permeability coefficient and adiabatic temperature rise value, the training set is 80% of the total data set, and the test set is 20% of the total data set; S402, obtaining the associated physical equation set, wherein the associated physical equation set includes the mechanical constitutive equation, the heat conduction equation and the chloride ion diffusion equation; S403, constructing an initial PINN model, determining the number of layers of the hidden layer and the number of neurons of each layer of the initial PINN model according to the highest order of the mechanical constitutive equation, the heat conduction equation and the chloride ion diffusion equation; S404, determining the residual of the mechanical constitutive equation, the heat conduction equation and the chloride ion diffusion equation based on the associated physical equation set and the initial PINN model to obtain the mechanical residual, the heat conduction residual and the chloride ion diffusion residual, obtaining the mechanical loss term, the heat conduction loss term and the chloride ion diffusion loss term based on the mean square error formula, the mechanical residual, the heat conduction residual and the chloride ion diffusion residual, determining the initial condition loss term and the boundary condition loss term, and obtaining the total loss function based on the mechanical loss term, the heat conduction loss term, the chloride ion diffusion loss term, the boundary condition loss term and the initial condition loss term, wherein the expression of the mean square error formula is: The expression of the mechanical loss term is: The expression of the heat conduction loss term is: The expression of the chloride ion diffusion loss term is: n is the number of all sample data in the training set, is the true value of the i th sample data in the training set, is the predicted value of the i th sample data in the training set, is the mechanical loss term, is the heat conduction loss term, is the chloride ion diffusion loss term, is the number of selected points in the region in the definition domain, and the expression of the initial condition loss term is: The expression of the boundary condition loss term is: wherein, is an initial condition loss term, is a boundary condition loss term, is a number of region sampling points of the initial condition, is a number of region sampling points of the boundary condition, is the i-th sample data in the total sample data, and an expression of the total loss function is: , is a weight of the mechanical loss term, is a weight of the heat conduction loss term, is a weight of the chloride ion diffusion loss term, is a weight of the initial condition loss term, is a weight of the boundary condition loss term; S405, training the initial PINN model based on the total loss function using a neural network Adam and L-BFGS optimizer hybrid strategy to obtain a target PINN model; S406, obtaining a test set, obtaining predicted values of all sample data of the test set based on the test set and the target PINN model, obtaining true values of the sample data, obtaining a root mean square error, a determination coefficient and a mean absolute error based on the predicted values, the true values, a root mean square error formula, a determination coefficient formula and a mean absolute error formula, when the determination coefficient is greater than a first preset threshold, the root mean square error is less than a second preset threshold, and the mean absolute error is less than a third preset threshold, the fitting precision and the prediction accuracy of the target PINN model meet the requirements, wherein an expression of the root mean square error formula is: , an expression of the determination coefficient is: , and an expression of the mean absolute error is: , RMSE is the root mean square error, z is a number of sample data in the test set, is a true value of the m-th sample data in the test set, is a predicted value of the m-th sample data in the test set, is the determination coefficient, is an average value of the true values of the sample data, MAE is the mean absolute error, z and m are both positive integers, and m is less than or equal to z; S407, constructing a target function group based on the target PINN model, wherein the target function group includes a compression resistance target function, a chloride ion permeability coefficient target function, an adiabatic temperature rise value target function and an economic cost target function, an expression of the compression resistance target function is: , an expression of the chloride ion permeability coefficient is: , an expression of the adiabatic temperature rise value target function is , and an expression of the economic cost target function is: , a regression function of the compressive strength of the three-grade concrete, a regression function of the chloride ion permeability coefficient of the three-grade concrete, a regression function of the adiabatic temperature rise of the three-grade concrete, a regression function of the economic cost of the three-grade concrete, X is a parameter variable of a mixing material, and the parameter variable of the mixing material includes a water-binder ratio, a cement dosage, a sand dosage, a gravel dosage, a fly ash dosage, a slag powder dosage, and a water reducing agent dosage, a water-binder ratio, a cement dosage, a sand dosage, a gravel dosage, a fly ash dosage, a slag powder dosage, a water reducing agent dosage, a unit mass cost price of the parameter variable of the mixing material, the i-th parameter variable of the mixing material, S408, constructing an initial constraint condition, and obtaining a target constraint condition based on the initial constraint condition, wherein an expression of the initial constraint condition is: , the i-th parameter variable of the mixing material, a lower limit of a standard design of the i-th parameter variable of the mixing material, an upper limit of the standard design of the i-th parameter variable of the mixing material, and an expression of the target constraint condition is: ; S409, performing multi-factor optimization on the target PINN model by using a non-dominated sorting genetic algorithm, and setting an iteration termination condition, wherein the iteration termination condition includes a size of an initial population, a maximum iteration number, and a convergence standard Pareto change rate; S410, generating the initial population based on the iteration termination condition, wherein the initial population includes a plurality of individuals, each individual is used to represent a group of parameter combinations of a mix proportion of the three-grade concrete, and the individual satisfies the target constraint condition; S411, performing fitness evaluation and non-dominated sorting on all individuals of the initial population based on the target function group, obtaining a fitness evaluation result and a non-dominated sorting result, determining a Pareto solution set through the fitness evaluation result and the non-dominated sorting result, determining an optimal mix proportion based on all Pareto solutions of the Pareto solution set, and outputting the optimal mix proportion.

[0063] Based on the complete technical solution of the application, sample data is obtained based on performance indexes including compressive strength, durability and crack resistance, a single compressive strength is avoided as an optimization target, at least one of the compressive strength, the durability and the crack resistance is taken as an optimization target for a Pareto solution set output by a target PINN model, the needs of the three-grade concrete applied to different severe environments are met; the target PINN model is obtained based on all sample data and associated physical equations, a partial differential equation is introduced as a physical constraint, the target PINN model has physical meaning, the target function set output based on the target PINN model conforms to the basic physical law, the dependence on data is reduced, and the model generalization ability is improved; in the mix proportion optimization problem, the target function set determined based on the target PINN model embedded with the physical equation constraint can avoid outputting unreasonable Pareto solutions, and can accelerate the optimization process to bring more accurate and reasonable optimization results. Through the complete technical solution of the application, the dependence of the target PINN model on sample data can be reduced, the physical meaning and interpretability of the target PINN model can be enhanced, and the prediction accuracy of the compressive strength, the durability and the crack resistance can be improved.

[0064] As Figure 2 shown, Figure 2 is a structural diagram of a three-grade concrete mix proportion optimization device provided by an embodiment of the application. The application further provides a three-grade concrete mix proportion optimization device, which comprises: The processor 501 can be implemented in a general-purpose central processing unit (CPU), a microprocessor, an application specific integrated circuit (ASIC), or one or more integrated circuits, and is used to execute related programs to implement the technical solutions provided by the embodiments of the application. The memory 502 can be implemented in the form of a read only memory (ROM), a static storage device, a dynamic storage device, or a random access memory (RAM). The memory 502 can store an operating system and other application programs. When the technical solutions provided by the embodiments of the application are implemented by software or firmware, the related program codes are stored in the memory 502 and are called and executed by the processor 501 to implement the three-grade concrete mix proportion optimization method of the embodiments of the application. The input / output interface 503 is used to realize information input and output. The communication interface 504 is configured to realize the communication interaction between the device and other devices, and can realize the communication through a wired manner (for example, a USB, a network cable and the like) or a wireless manner (for example, a mobile network, WIFI, Bluetooth and the like). The bus 505 is configured to transmit information between various components (for example, the processor 501, the memory 502, the input / output interface 503 and the communication interface 504) of the device. The processor 501, the memory 502, the input / output interface 503 and the communication interface 504 are connected to each other through the bus 505.

[0065] The embodiment of the present application further provides an electronic device, which comprises the three-grade concrete mixing proportion optimization device.

[0066] The embodiment of the present application further provides a storage medium, which is a computer readable storage medium, and the storage medium stores a computer program, and the computer program is executed by a processor to realize the three-grade concrete mixing proportion optimization method.

[0067] The memory is a non-transient computer readable storage medium, and can be used to store a non-transient software program and a non-transient computer executable program. In addition, the memory can include a high-speed random access memory, and can further include a non-transient memory, for example, at least one magnetic disk storage device, a flash memory device or other non-transient solid-state memory device. In some embodiments, the memory can optionally include a memory remotely arranged relative to the processor, and the remote memory can be connected to the processor through a network. Examples of the network include but are not limited to the Internet, an intranet, a local area network, a mobile communication network and a combination thereof. The device embodiment described above is only schematic, and the units described as separate components can be or can not be physically separated, and can be located in one place or can be distributed on multiple network units. According to actual needs, part or all of the modules can be selected to realize the purpose of the embodiment.

[0068] As will be appreciated by one of ordinary skill in the art, all or some of the steps, systems, etc. in the above-disclosed methods can be embodied in software, firmware, hardware, and / or suitable combinations thereof. Some or all of the physical components can be implemented with software executed by a processor, such as a central processing unit, a digital signal processor, or microprocessor, or can be implemented as hardware, or as an integrated circuit, such as an application- specific integrated circuit. Such software can be distributed on computer readable media, which can comprise computer storage media (or non-transitory media), and communication media (or transitory media). As is well known to those of ordinary skill in the art, computer storage media includes both volatile and nonvolatile, removable and non-removable media implemented in any method or technology for storage of information such as computer readable instructions, data structures, program modules or other data. Computer storage media includes, but is not limited to, RAM, ROM, EEPROM, flash memory or other memory technology, CD-ROM, digital versatile disks (DVD) or other optical disk storage, magnetic cassettes, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other medium which can be used to store the desired information and which can be accessed by a computer. Further, as is well known to those of ordinary skill in the art, communication media typically embodies computer readable instructions, data structures, program modules, or other data in a modulated data signal, such as carrier waves or other transport mechanisms, and includes any information delivery media.

[0069] The above description is that of the preferred embodiments of the application. Various modifications and changes can be made thereto without departing from the spirit and scope of the application, which is to be given the broadest interpretation of the laws.

Claims

1. A method for optimizing the mix proportion of three-graded concrete, characterized in that, include: Multiple sets of sample data for three-grade concrete were obtained, and a training set was obtained based on all the sample data. Obtain the associated physical equation set and the initial PINN model, determine the total loss function based on the associated physical equation set and the training set, and train the initial PINN model based on the total loss function to obtain the target PINN model, wherein the associated physical equation set includes at least one physical equation related to three-grade concrete. Based on the target PINN model, a set of objective functions is constructed, and objective constraints are determined. Based on the objective constraints and the target PINN model, a Pareto solution set of the objective function set is generated using a non-dominated sorting genetic algorithm. Based on the Pareto solution set, the optimal mix proportion of the three-grade concrete is determined, and the optimal mix proportion is output. The Pareto solution set includes multiple Pareto solutions.

2. The method for optimizing the mix proportion of three-graded concrete according to claim 1, characterized in that, Obtain the associated physical equations and the initial PINN model, and determine the total loss function based on the associated physical equations and the training set, including: Obtain the associated physical equation set, which includes a mechanical constitutive equation, a heat conduction equation, and a chloride ion diffusion equation. The mechanical constitutive equation, the heat conduction equation, and the chloride ion diffusion equation are expressed in partial differential equation form. The expression for the mechanical constitutive equation is: The expression for the heat conduction equation is: The expression for the chloride ion diffusion equation is as follows: , Let ρ be the stress, f be the volume force per unit volume, T be the temperature, ρ be the density, c be the specific heat capacity, k be the thermal conductivity, Q be the heat source term, t be the time, D be the chloride ion permeability coefficient, and C be the chloride ion concentration. The initial PINN model is constructed, and the number of hidden layers and the number of neurons in each layer are determined based on the highest order of the mechanical constitutive equation, the heat conduction equation, and the chloride ion diffusion equation. The total loss function is obtained based on the associated physical equations and the initial PINN model.

3. The method for optimizing the mix proportion of three-graded concrete according to claim 2, characterized in that, The total loss function is obtained based on the associated physical equations and the initial PINN model, including: Based on the associated physical equations and the initial PINN model, the residuals of the mechanical constitutive equation, the heat conduction equation, and the chloride ion diffusion equation are determined respectively, resulting in the mechanical residual, the heat conduction residual, and the chloride ion diffusion residual. Based on the mean square error formula, the mechanical residual, the thermal conduction residual, and the chloride ion diffusion residual, the mechanical loss term, thermal conduction loss term, and chloride ion diffusion loss term are obtained respectively. The expression for the mean square error formula is: The expression for the mechanical loss term is: The expression for the heat conduction loss term is: The expression for the chloride ion diffusion loss term is: Where n is the total number of all sample data in the training set. The true value of the i-th sample data in the training set. The predicted value of the i-th sample data in the training set. For the aforementioned mechanical loss term, For the heat conduction loss term, This refers to the chloride ion diffusion loss term. n and i represent the number of points selected within the domain, where n and i are both positive integers. Determine the initial condition loss term and the boundary condition loss term, wherein the expression for the initial condition loss term is: The expression for the boundary condition loss term is: ,in, For the initial condition loss term, The boundary condition loss term, The number of points selected in the region as the initial condition. The number of points to select for the boundary conditions region. The i-th sample data in all the sample data; The total loss function is obtained based on the mechanical loss term, the thermal conduction loss term, the chloride ion diffusion loss term, the boundary condition loss term, and the initial condition loss term, wherein the expression of the total loss function is: , The weight of the mechanical loss term, The weight of the heat conduction loss term, The weight of the chloride ion diffusion loss term. The weights of the initial conditional loss term, The weights of the boundary condition loss term.

4. The method for optimizing the mix proportion of three-graded concrete according to claim 1, characterized in that, Multiple sets of sample data for three-grade concrete were obtained, and a training set was obtained based on all the sample data, including: Obtain the mix proportion materials and performance indicators of three-grade concrete, and obtain all the sample data based on the mix proportion materials and performance indicators, wherein the performance indicators include concrete compressive strength, chloride ion permeability coefficient and adiabatic temperature rise value. Data cleaning is performed on all the sample data, and then normalization is performed on all the sample data based on a normalization formula to obtain the total dataset, wherein the normalization formula is: , Normalized value For the i-th sample data in all the sample data, For the j-th sample data in all the sample data, The minimum value among all the sample data. The maximum value among all the sample data is e, where e is the number of all the sample data, and both e and j are positive integers. The total dataset is divided into the training set and the test set.

5. The method for optimizing the mix proportion of three-graded concrete according to claim 4, characterized in that, After training the initial PINN model based on the total loss function to obtain the target PINN model, the process further includes: Obtain the test set, and based on the test set and the target PINN model, obtain the predicted values ​​of all the sample data in the test set; Obtain the true values ​​of all sample data in the test set. Based on all predicted values ​​in the test set, all true values ​​in the test set, the root mean square error formula, the coefficient of determination formula, and the mean absolute error formula, obtain the root mean square error, the coefficient of determination, and the mean absolute error, respectively. The expression for the root mean square error formula is: The expression for the coefficient of determination is: The expression for the mean absolute error is: RMSE is the root mean square error, and z is the number of sample data in the test set. The true value of the m-th sample data in the test set. The predicted value of the m-th sample data in the test set. The determination coefficient is... z is the average of the true values ​​of the sample data, MAE is the mean absolute error, z and m are both positive integers, and m is less than or equal to z; When the coefficient of determination is greater than a first preset threshold, the root mean square error is less than a second preset threshold, and the mean absolute error is less than a third preset threshold, the fitting accuracy and prediction accuracy of the target PINN model meet the requirements.

6. The method for optimizing the mix proportion of three-graded concrete according to claim 4, characterized in that, Based on the target PINN model, a set of objective functions is constructed, and the objective constraints are determined, including: The objective function set is constructed based on the target PINN model, wherein the objective function set includes a pressure resistance objective function, a chloride ion permeability coefficient objective function, an adiabatic temperature rise objective function, and an economic cost objective function. The expression of the pressure resistance objective function is as follows: The expression for the chloride ion permeability coefficient is: The expression for the objective function of the adiabatic temperature rise is: The expression for the economic cost objective function is: , The regression function for the compressive strength of three-graded concrete is given. This is a regression function for the chloride ion permeability coefficient of three-grade concrete. The regression function for the adiabatic temperature rise of three-grade concrete. Let X be the regression function for the economic cost of three-grade concrete, and let X be the parameter variables of the mix proportion materials, including water-cement ratio, cement content, sand content, crushed stone content, fly ash content, slag powder content, and water-reducing agent content. The water-to-binder ratio is [the specified value]. The amount of cement used is [amount]. The amount of sand used is... The amount of crushed stone used is... The amount of fly ash used is [amount]. The amount of slag powder used is... The amount of water-reducing agent used is... The unit mass cost price of the material parameters in the mix proportion. Let i be the parameter variable of the i-th material in the mix proportion; Construct initial constraints, wherein the expressions for the initial constraints are: , Let i be the parameter variable of the i-th material in the mix proportion. This represents the lower limit of the specification design for the parameter variables of the i-th material in the mix proportion. The upper limit of the parameter variables for the i-th material in the mix proportion, as specified in the standard design; The target constraint is obtained based on the initial constraint, wherein the expression of the target constraint is: .

7. The method for optimizing the mix proportion of three-graded concrete according to claim 1, characterized in that, Based on the objective constraints and the objective PINN model, a Pareto solution set for the objective function set is generated using a non-dominated sorting genetic algorithm. Based on the Pareto solution set, the optimal mix proportion of the three-grade concrete is determined, and the optimal mix proportion is output, including: The target PINN model is optimized using the non-dominated sorting genetic algorithm, and the iteration termination condition is set, which includes the initial population size, the maximum number of iterations, and the convergence standard Pareto rate of change. Based on the iteration termination condition, the initial population is generated, wherein the initial population includes multiple individuals, each individual representing a set of parameter combinations of the mix proportion of three-grade concrete, and the individual satisfies the target constraint condition. Based on the objective function set, fitness evaluation and non-dominated sorting are performed on all individuals in the initial population to obtain fitness evaluation results and non-dominated sorting results. The Pareto solution set is determined by the fitness evaluation results and the non-dominated sorting results. The optimal mix ratio is determined based on all Pareto solutions in the Pareto solution set, and the optimal mix ratio is output.

8. A three-grade concrete mix proportion optimization device, characterized in that, It includes at least one control processor and a memory for communicatively connecting to the at least one control processor; the memory stores instructions executable by the at least one control processor, which, when executed by the at least one control processor, enable the at least one control processor to perform the three-grade concrete mix proportion optimization method as described in any one of claims 1 to 7.

9. An electronic device, characterized in that, It includes the three-grade concrete mix proportion optimization device as described in claim 8.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer-executable instructions for causing a computer to perform the three-grade concrete mix proportion optimization method as described in any one of claims 1 to 7.

Images