Foamed light soil internal damage process prediction method and system

Abstract

The invention relates to the technical field of testing or analyzing materials by means of measuring chemical or physical properties of the materials, in particular to a method and a system for predicting an internal damage process of foam light soil, and the method comprises the following steps: constructing a linear and elastic constitutive physical model of the foam light soil based on a near-field dynamics theory; defining a force-displacement relation and a damage evolution criterion of a non-local key; executing fusion of the linear and elastic constitutive physical model and a machine learning network, and determining a hybrid model; and obtaining state parameters of the foam light soil, and executing spatio-temporal evolution of a prediction damage field based on the hybrid model so as to realize prediction of the internal damage process of the foam light soil. According to the method, multi-scale prediction of the internal damage condition of the foam light soil pore structure is realized.

Description

Technical Field

[0001] The present invention relates to the technical field of testing or analyzing materials by measuring the chemical or physical properties of materials, and in particular to a method for predicting the internal damage process of foamed lightweight soil, and also to a prediction system. Background Art

[0002] Foamed lightweight soil is a lightweight engineering material made by mixing a foaming agent with a cement-based material, and has excellent properties such as loose pores, light weight and high strength, good fluidity, and environmental friendliness. In recent years, foamed lightweight soil has been widely used in the engineering construction of infrastructure such as roads, railways, and airports. However, due to the complex pore structure characteristics inside the foamed lightweight soil, it has become a difficult problem to predict the internal structural damage evolution process of the foamed lightweight soil during its life cycle.

[0003] There is a Chinese patent application for invention with the publication number CN119510455A, the publication date of February 25, 2025, and the title of a non-destructive testing method for the internal pore structure of foamed lightweight soil, which discloses the following technical solution: Take a cube-shaped foamed lightweight soil specimen with a side length of 40 mm, place the cured specimen on the sample stage, and scan the specimen once around with an X-CT scanner. The X-ray passes through different layers of the specimen, and the detector receives the information attenuated by different layers. After being converted by software, it is input into a computer for reconstruction to obtain a grayscale image; Use Image-ProPlus analysis software to process the grayscale image, automatically identify the grayscale image threshold, and use the WatershedSplit algorithm built into the Image-ProPlus analysis software to re-segment the automatically identified grayscale image, and then perform binary processing on the segmented image.

[0004] Although the foregoing technology can achieve non-destructive testing of the porosity, pore size distribution, pore shape factor, and average pore size inside the foamed lightweight soil, it cannot provide an effective solution for predicting the internal structural damage evolution of the foamed lightweight soil. Summary of the Invention

[0005] The inventors have found through research that the peridynamics theory is a new non-local continuum mechanics theory system, which uses integral formulas in the control equations, rather than partial differential equations, to describe the response of continuous or discontinuous media under stress, making peridynamics very suitable for simulating the damage and fracture processes of materials without complex numerical processing at discontinuous surfaces or crack tips. Therefore, by using the peridynamics method and the deep learning algorithm, it is possible to predict the internal damage process of foamed lightweight soil.

[0006] The object of the present invention is to provide a method and system for predicting the internal damage process of foamed lightweight soil, which solves the technical problem that the prior art cannot provide a method for reasonably predicting the internal damage of foamed lightweight soil by integrating the linear and elastic constitutive physical model of foamed lightweight soil with a machine learning network; meanwhile, it solves the technical problem that the prior art cannot provide a system for reasonably predicting the internal damage of foamed lightweight soil.

[0007] According to one aspect of the present invention, there is provided a method for predicting the internal damage process of foamed lightweight soil, which is executed by a processor and includes: constructing a linear and elastic constitutive physical model of foamed lightweight soil based on the peridynamics theory, and executing the definition of the force-displacement relationship and damage evolution criterion of non-local bonds; executing the integration of the linear and elastic constitutive physical model and the machine learning network to determine a hybrid model; obtaining the state parameters of foamed lightweight soil, and based on the hybrid model, executing the prediction of the spatio-temporal evolution of the damage field to achieve the prediction of the internal damage process of foamed lightweight soil.

[0008] In some embodiments, the process of constructing the internal damage physical model of foamed lightweight soil based on the peridynamics theory and executing the definition of the force-displacement relationship and damage evolution criterion of non-local bonds is as follows: executing the determination of the linear and elastic constitutive physical model of foamed lightweight soil with pore structure; based on the linear and elastic constitutive physical model of foamed lightweight soil, executing the strain rate specification to simulate the uniaxial compression process of the matrix material of foamed lightweight soil; executing the determination of the motion-time equation of the matrix particles of foamed lightweight soil; solving based on the motion-time equation of the matrix particles of foamed lightweight soil, recording the stress-strain curve of foamed lightweight soil at each time step, and performing damage analysis and calculation.

[0009] In some embodiments, the process of executing the determination of the linear and elastic constitutive physical model of foamed lightweight soil with pore structure is as follows: based on explicit modeling and with the porosity constraint, generating a number of independent pore units; based on the independent pore units, executing the determination of the porosity, the sphericity of the pore units, the model size of foamed lightweight soil, and the particle spacing; uniformly discretizing based on the discrete spacing, and storing the coordinates of the matrix particles of foamed lightweight soil; traversing all the matrix particles, determining other particles with influencing results, recording the positions of the particles with influencing results, and executing the storage of the corresponding arrays.

[0010] In some embodiments, the process of executing the determination of the motion-time equation of the matrix particles of foamed lightweight soil is as follows: calculating the interaction between the matrix particles based on the bond-based peridynamics theory formula; performing numerical integration on the forces between the matrix particles in the peridynamic neighborhood to obtain the internal forces received by the matrix particles, so as to establish the time-motion equation of the matrix particles of foamed lightweight soil.

[0011] In some embodiments, the process of specifying the strain rate based on the linear and elastic constitutive physical model of foamed lightweight soil to simulate the uniaxial compression process of the foamed lightweight soil matrix material is as follows: Specify the strain rate for the top and bottom surfaces of the model, and execute the limit that the external force density b = 0.

[0012] In some embodiments, the process of solving the motion-time equation of the foamed lightweight soil matrix particles, recording the stress-strain curve of the foamed lightweight soil at each time step, and performing damage analysis and calculation is as follows: Divide the foamed lightweight soil matrix particles with a specified strain rate into n time steps; for any time step, execute the combination of the external force density and the time-motion equation, and solve the acceleration corresponding to the time step; according to the explicit time integration method, calculate the velocity and displacement of the matrix particles at the time step, and record the stress and strain data of all time steps; calculate the bond elongation amount between each matrix particle corresponding to the n + 1 time step and other particles in its neighborhood; repeat the judgment of whether the time integration is completed until the time integration is completed; store the calculation results of all time steps in the result file; perform post-processing on the result file, and slice the linear and elastic constitutive physical model to analyze the internal damage process of the foamed lightweight soil.

[0013] In some embodiments, the process of executing the fusion of the internal damage physical model and the machine learning network to determine the hybrid model is as follows: Based on the data and images obtained from the constructed internal damage physical model of the foamed lightweight soil, fuse the internal damage data of the foamed lightweight soil obtained from multi-scale tests, and optimize and generate a damage evolution data set; based on the optimized and generated damage evolution data set, construct a physics-informed neural network learning model; execute the fusion of the internal damage physical model and the physics-informed neural network learning model; determine the physical constraint loss function to optimize the model parameters, and train the physics-informed neural network learning model to determine the hybrid model, where during the training process, the weights and biases of the physics-informed neural network learning model are adaptively adjusted by minimizing the comprehensive loss function.

[0014] According to another aspect of the present invention, a prediction system for the internal damage process of foamed lightweight soil is provided. The system includes a processor, and further includes: a construction module, which is used to construct an internal damage physical model of foamed lightweight soil based on the peridynamics theory and execute the definition of the force-displacement relationship and damage evolution criterion of the non-local bond; a fusion module, which is used to execute the fusion of the internal damage physical model and the machine learning network to determine the hybrid model; an execution module, which is used to obtain the state parameters of the foamed lightweight soil and execute the prediction of the spatio-temporal evolution of the damage field based on the hybrid model to realize the prediction of the internal damage process of the foamed lightweight soil.

[0015] In some embodiments, the construction module, the fusion module, and the execution module are all data-connected to the processor.

[0016] According to one or more technical solutions adopted in the present invention, the beneficial effects that can be achieved are as follows: By establishing a linear and elastic constitutive model of foamed lightweight soil using peridynamics and combining it with machine learning, a physics-informed neural network learning model is constructed. Through the training and optimization of the physics-informed neural network learning model, a hybrid model for predicting the internal damage process of foamed lightweight soil is obtained, realizing multi-scale prediction of the internal damage conditions of the pore structure of foamed lightweight soil. Description of the Drawings

[0017] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0018] Figure 1 is the flowchart of the prediction method of the present invention;

[0019] Figure 2 is the structural diagram of the prediction system of the present invention;

[0020] Figure 3 is the schematic diagram of the principle of simulating the internal damage of foamed lightweight soil corresponding to the embodiment of the present invention;

[0021] Figure 4 is the flowchart of the peridynamics calculation corresponding to the embodiment of the present invention;

[0022] Figure 5 is the schematic diagram of the internal structure slice of the foamed lightweight soil of the present invention. Detailed Embodiments

[0023] The following will combine the drawings in the embodiments of the present invention Figures 1-5 to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, rather than all embodiments.

[0024] Application Overview:

[0025] The peridynamics theory is a new non-local continuum mechanics theory system. By using the integral formula in the control equation to describe the response of continuous or discontinuous media under stress, peridynamics is very suitable for simulating the damage and fracture processes of materials without complex numerical processing at discontinuous surfaces or crack tips.

[0026] Meanwhile, due to its excellent self-learning ability and powerful adaptation and fitting ability, machine learning technology can be used to predict data and images based on a dataset. Therefore, by combining peridynamics and machine learning, the prediction of the internal damage process of foamed lightweight soil can be achieved.

[0027] Embodiment 1

[0028] Figure 1 The following is a flowchart of a method for predicting the internal damage process of foamed lightweight soil provided by an embodiment of the present invention. This method is executed by a processor, which can be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), off-the-shelf programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc.; the general-purpose processor can be a microprocessor or any conventional processor, etc.

[0029] The method specifically includes:

[0030] Construct a linear and elastic constitutive physical model of foamed lightweight soil based on the peridynamics theory, and define the force-displacement relationship and damage evolution criterion of non-local bonds. Specifically, in some possible implementation manners, determine the linear and elastic constitutive physical model of foamed lightweight soil with pore structure. By the random placement method in the explicit modeling method, combined with the porosity constraint, generate a number of independent pore units. That is, determine the size of the foamed lightweight soil specimen as 6mm×6mm×6mm. Under the porosity constraint, randomly generate multiple independent pore units and allow the pores to overlap. Further, based on the independent pore units, determine the porosity, the sphericity of the pore units, the size of the foamed lightweight soil model, and the particle spacing. It should be noted that the porosity of the foamed lightweight soil in this embodiment is 10% - 70%, the sphericities of the pore units are 0.6, 0.8, and 1.0 respectively, and the particle spacing is set to 0.05mm. Further, perform uniform discretization based on the discrete spacing and store the coordinates of the particles of the foamed lightweight soil matrix. It can be understood that in other possible implementation environments, the numerical values in this embodiment can be adaptively adjusted and modified. Further, traverse all the matrix particles, determine other particles with influencing results, record the positions of the particles with influencing results, and store the corresponding arrays. Specifically, traverse all the particles. Assume that the effective set is a sphere centered at x with a radius of ζ. The effective set in this embodiment is defined as 3 times the particle spacing. Determine other particles that will interact with this particle according to the particles within the so-called effective set range. The relative position of the interacting particles is defined as bond ξ, and its expression is:

[0031] ξ = x' - x

[0032] And record the positions of the corresponding particles and store them in the corresponding arrays.

[0033] In some possible implementation manners, the process of determining the motion-time equation of the particles of the foamed lightweight soil matrix is as follows: Based on the bond-based peridynamics theory formula, calculate the interaction between the matrix particles; perform numerical integration on the forces between the matrix particles in the peridynamic neighborhood to obtain the internal forces received by the matrix particles, so as to establish the time-motion equation of the particles of the foamed lightweight soil matrix. Specifically, perform numerical integration on the forces between the particles in the peridynamic neighborhood to obtain the internal forces received by the particles. Then, assume that the external force density received by the particles is b, and establish the time-motion equation of the particles of the foamed lightweight soil matrix:

[0034]

[0035] Among them, ρ is the particle density, ü is the acceleration of the particle, ζ is the peridynamic neighborhood range of particle x, and T is the vector field describing the mutual force between points. It should be noted that each particle interacts with the particles whose distance from it is less than ζ, and the integration region is set as an ellipsoid with a radius of ζ.

[0036] In some possible embodiments, based on the linear and elastic constitutive physical model of foamed lightweight soil, strain rate specification is performed to simulate the uniaxial compression process of the foamed lightweight soil matrix material. Specifically, strain rates are specified for the top and bottom surfaces of the model, and the external force density b = 0 is enforced. It should be noted that a strain rate of 20 s -1 is specified for the top and bottom surfaces of each linear and elastic constitutive physical model of foamed lightweight soil, that is, the compression rate. The deformation of other surfaces is not restricted to simulate uniaxial compression. Since boundary displacement constraints are imposed on the linear and elastic constitutive physical model of foamed lightweight soil, the load is directly transmitted to the interior of the material through the boundary displacement constraints, and at this time the external force density b = 0.

[0037] In some possible embodiments, the motion-time equation of the foamed lightweight soil matrix particles is solved, and the stress-strain curve of the foamed lightweight soil at each time step is recorded, and the process of damage analysis calculation is as follows: for the foamed lightweight soil matrix particles with a specified strain rate applied, n time steps are divided; for any time step, the combination of the external force density and the time-motion equation is performed, and the acceleration corresponding to the time step is solved; according to the explicit time integration method, the velocity and displacement of the matrix particles at the time step are calculated, and the stress and strain data of all time steps are recorded; the bond elongation between each matrix particle and other particles in its neighborhood at the n + 1 time step is calculated; the judgment of whether the time integration is completed is repeatedly executed until the time integration is completed; the calculation results of all time steps are stored in the result file; post-processing is performed on the result file, and the linear and elastic constitutive physical model is sliced to analyze the internal damage process of the foamed lightweight soil. Preferably, the foamed lightweight soil matrix particles with a specified strain rate applied are divided into n t time steps. At the n t time step, the calculated external force density b is substituted into the time-motion equation, and the acceleration at the n t time step is solved. The velocity and displacement of the particles at the n t + 1 time step are calculated using the explicit time integration method, and the stress and strain data of each time step are recorded. The expression of the explicit time integration method is as follows:

[0038]

[0039] where △t is the time step size.

[0040] Next, the bond elongation between each particle and other particles in its neighborhood at the n t + 1 time step is calculated, and the calculation expression is as follows:

[0041] η = u(x′, t) - u(x, t)

[0042] Among them, η is the relative displacement of the particle before and after deformation; u is the displacement vector field;

[0043] Furthermore, according to the obtained bond elongation, the fracture situation of the bond is judged, and the damage value D(x,t) of the particle is calculated. The calculation expression is as follows:

[0044]

[0045]

[0046] Among them, S C is the critical tensile value between particles. μ = 1 indicates that the bond has not broken, and μ = 0 indicates that the bond has broken; It should be noted that the calculation expression of S C is:

[0047]

[0048] Among them, G f is the fracture energy of the material, taking 1.75 J / m 2 ; K and G are the bulk modulus and shear modulus respectively.

[0049] Furthermore, it is judged whether the time integration is completed. If not, the above process is repeated until the time integration is completed; Furthermore, the calculation results of all time steps are stored in the result file, and after the calculation is completed, the result file is post-processed and the internal damage process of the foamed lightweight soil material is analyzed.

[0050] Execute the fusion of the linear and elastic constitutive physical model and the machine learning network to determine the hybrid model. Specifically, based on the data and images obtained from the established internal damage physical model of the foamed lightweight soil, fuse the internal damage data of the foamed lightweight soil obtained from the multi-scale test, and optimize and generate the damage evolution data set; Based on the optimized generated damage evolution data set, construct a physics-informed neural network learning model; Execute the fusion of the internal damage physical model and the physics-informed neural network learning model; Determine the physical constraint loss function to optimize the model parameters, and train the physics-informed neural network learning model to determine the hybrid model. Among them, during the training process, the weights and biases of the physics-informed neural network learning model are adaptively adjusted by minimizing the comprehensive loss function.

[0051] It should be noted that to construct a physics-informed neural network learning model, namely PINN, the established linear and elastic constitutive physical model of foamed lightweight soil is embedded in the machine learning network. The PINN learning model includes an input layer, an output layer, and a hidden layer. The hidden layer in the PINN learning model is a multi-layer fully connected network layer, and its activation function is ReLU or Tanh. By inputting the parameter characteristics of the internal pore structure of foamed lightweight soil through the input layer, the hidden layer captures the non-linearity of the input data, and finally outputs the internal damage characteristics of foamed lightweight soil at the output layer. Furthermore, a physical constraint loss function is constructed to optimize the model parameters, and the aforementioned PINN learning model is trained. The weights and biases of the PINN learning model are adaptively adjusted through the loss function. Among them, the physical constraint loss function includes physical residual loss, boundary condition loss, and data loss.

[0052] In a preferred embodiment, it further includes: the physical constraint loss function is the weighted sum of physical residual loss, boundary condition loss, and data loss, and the function expression is as follows:

[0053] μ = λ phys μ phys +λ bc μ bc +λ data μ data

[0054] Wherein, μ represents the total physical constraint loss function; μ phys represents the physical residual loss; μ bc represents the boundary condition loss; μ data represents the data loss; λ phys 、λ bc and λ data respectively represent the weight coefficients of physical residual loss, boundary condition loss, and data loss, and μ phys 、μ bc 、μ data expressions are as follows respectively:

[0055]

[0056]

[0057]

[0058] Wherein, N phys represents the number of sampling points in the domain; (x i , t i ) represents the points randomly sampled in the space-time domain; N bc represents the number of boundary sampling points; represents the coordinates on the boundary; Represents the true damage value on the boundary; u NN Represents the damage value predicted by the PINN learning model; u obs Represents the actual damage value measured by numerical simulation.

[0059] In a preferred embodiment, it further includes: selecting an optimizer to train the constructed PINN learning model, calculating the physical residual loss, boundary condition loss, and data loss respectively, and then obtaining the total physical constraint loss function. The weights of each loss term are adjusted using dynamic loss weights, and the calculation expression is as follows:

[0060]

[0061] Furthermore, update the total loss, and the expression is as follows:

[0062]

[0063] Then, calculate the gradient of the total physical constraint loss function with respect to the parameters of the PINN learning model through an automatic differentiation tool, and the expression is as follows:

[0064]

[0065] Furthermore, use an optimizer, such as Adam, to update the parameters of the PINN learning model. When the above training steps make the loss function converge, the PINN model training is completed, and the optimal parameters of the PINN learning model are obtained. The specific expression is:

[0066]

[0067] where η is the learning rate, is the gradient of the loss function with respect to the parameters.

[0068] Obtain the state parameters of the foamed lightweight soil, and perform the spatio-temporal evolution of the predicted damage field based on the hybrid model to realize the prediction of the internal damage process of the foamed lightweight soil. Specifically, input the state parameters of the foamed lightweight soil measured in real time, such as the internal pore structure characteristics, such as porosity, sphericity, etc. Generate spatial grid points (x, y) in the PINN model, with a shape of (Nx, Ny). The relevant calculation expressions are as follows:

[0069] x = [x1, x2,... x Nx , y = [y1, y2,... y Nx

[0070] X, Y = meshgrid(x, y)

[0071] ​Furthermore, the internal structure damage prediction and visualization output of the foamed lightweight soil are obtained through processing by the PINN learning model; through the foregoing content, the PINN learning model can predict and output the spatio-temporal evolution process of the internal damage of the foamed lightweight soil.

[0072] Embodiment 2

[0073] Based on the same inventive concept as the prediction method for the internal damage process of the foamed lightweight soil in the foregoing Embodiment 1, as Figure 2 shown, the present invention also provides a prediction system for the internal damage process of the foamed lightweight soil, and the system includes a processor. The processor may be a Central Processing Unit (CPU), or may also be other general-purpose processors, Digital Signal Processors (DSPs), Application Specific Integrated Circuits (ASICs), Field-Programmable Gate Arrays (FPGAs) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc.; the general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.

[0074] Exemplarily, the prediction method for the internal damage process of the foamed lightweight soil may be divided into one or more modules, and one or more modules are stored in the memory and executed by the processor to complete the present invention. One or more modules may be a series of computer program instruction segments capable of completing specific functions, and the instruction segments are used to describe the execution process of the computer program. For example, the computer program may be divided into a construction module, a fusion module, and an execution module, and the specific functions of each module are as follows: the construction module is used to construct the internal damage physical model of the foamed lightweight soil based on the peridynamics theory and execute the definition of the force-displacement relationship and damage evolution criterion of the non-local bond; the fusion module is used to execute the fusion of the internal damage physical model and the machine learning network to determine the hybrid model; the execution module is used to obtain the state parameters of the foamed lightweight soil and execute the prediction of the spatio-temporal evolution of the damage field based on the hybrid model to realize the prediction of the internal damage process of the foamed lightweight soil.

[0075] In some possible implementation manners, the construction module, the fusion module, and the execution module are all data-connected to the processor.

[0076] The specific example of the prediction method for the internal damage process of the foamed lightweight soil in the foregoing First Embodiment is equally applicable to the prediction system for the internal damage process of the foamed lightweight soil in this embodiment. Through the foregoing detailed description of the prediction method for the internal damage process of the foamed lightweight soil, those skilled in the art can clearly know the prediction system for the internal damage process of the foamed lightweight soil in this embodiment. Therefore, for the sake of brevity of the specification, it will not be elaborated herein again.

[0077] The foregoing has shown and described the basic principles, main features and advantages of the present invention. For those skilled in the art, it is obvious that the present invention is not limited to the details of the above-described exemplary embodiments, and without departing from the spirit or basic characteristics of the present invention, the present invention can be implemented in other specific forms. Therefore, in any regard, the embodiments should be regarded as exemplary and non-limiting. The scope of the present invention is defined by the appended claims rather than the above description. Therefore, all changes falling within the meaning and scope of the equivalent elements of the claims are intended to be included in the present invention. Any reference signs in the claims should not be construed as limiting the claimed invention.

[0078] In addition, it should be understood that although this specification is described according to embodiments, not every embodiment only contains an independent technical solution. This narrative manner of the specification is only for clarity. Those skilled in the art should regard the specification as a whole, and the technical solutions in each embodiment can also be appropriately combined to form other embodiments that can be understood by those skilled in the art.

Claims

1. A prediction method for the internal damage process of foamed lightweight soil, which is executed by a processor, characterized in that Including: Construct a linear and elastic constitutive physical model of foamed lightweight soil based on the peridynamics theory, and execute the force-displacement relationship and damage evolution criterion for defining non-local bonds; Execute the fusion of the linear and elastic constitutive physical model and the machine learning network to determine the hybrid model; Obtain the state parameters of the foamed lightweight soil, and execute the prediction of the spatio-temporal evolution of the damage field based on the hybrid model to realize the prediction of the internal damage process of the foamed lightweight soil.

2. The method according to claim 1, characterized in that, The process of constructing the internal damage physical model of the foamed lightweight soil based on the peridynamics theory and executing the force-displacement relationship and damage evolution criterion for defining non-local bonds is as follows: Execute the determination of the linear and elastic constitutive physical model of the foamed lightweight soil with pore structure; Based on the linear and elastic constitutive physical model of the foamed lightweight soil, execute the strain rate specification to simulate the uniaxial compression process of the matrix material of the foamed lightweight soil; Execute the determination of the motion-time equation of the matrix particles of the foamed lightweight soil; Solve based on the motion-time equation of the matrix particles of the foamed lightweight soil, record the stress-strain curves of the foamed lightweight soil at each time step, and perform damage analysis and calculation.

3. The method according to claim 2, wherein The process of executing the determination of the linear and elastic constitutive physical model of the foamed lightweight soil with pore structure is as follows: Based on explicit modeling, cooperate with the porosity constraint to generate a number of independent pore units; Based on the independent pore units, execute the determination of the porosity, the sphericity of the pore units, the size of the foamed lightweight soil model, and the particle spacing; Perform uniform discretization based on the discrete spacing and store the coordinates of the matrix particles of the foamed lightweight soil; Traverse all matrix particles, determine other particles with influencing results, record the positions of the particles with influencing results, and execute the storage of the corresponding arrays.

4. The method according to claim 2, characterized in that, The process of executing the determination of the motion-time equation of the matrix particles of the foamed lightweight soil is as follows: Based on the formula of the bond-based peridynamics theory, calculate the interaction between matrix particles; Perform numerical integration on the forces between matrix particles in the peridynamic neighborhood to obtain the internal forces received by the matrix particles, so as to establish the time-motion equation of the matrix particles of the foamed lightweight soil.

5. The method according to claim 2, wherein The process of executing the strain rate specification based on the linear and elastic constitutive physical model of the foamed lightweight soil to simulate the uniaxial compression process of the matrix material of the foamed lightweight soil is as follows: Specify the strain rate on the top and bottom surfaces of the model, and execute the limitation of the external force density b = 0.

6. The method according to claim 2, characterized in that, The process of solving based on the motion-time equation of the matrix particles of the foamed lightweight soil, recording the stress-strain curves of the foamed lightweight soil at each time step, and performing damage analysis and calculation is as follows: Perform n time step divisions on the matrix particles of the foamed lightweight soil with the specified strain rate applied; For any time step, execute the combination of the external force density and the time-motion equation, and solve the acceleration corresponding to the time step; According to the explicit time integration method, calculate the velocity and displacement of the matrix particles at the time step, and record the stress and strain data at all time steps; Calculate the bond elongation between each matrix particle corresponding to the n+1 time step and other particles in its neighborhood; Repeat the execution of judging whether the time integration is completed until the time integration is completed; For the calculation results at all time steps, execute the storage to the result file; Post-process the result file and slice the linear and elastic constitutive physical models to analyze the internal damage process of the lightweight foamed soil.

7. The method according to claim 1, wherein The process of fusing the internal damage physical model with the machine learning network to determine the hybrid model is as follows: Based on the data and images obtained from the established internal damage physical model of the lightweight foamed soil, fuse the internal damage data of the lightweight foamed soil obtained from multi-scale tests, and optimize and generate a damage evolution data set; Based on the optimized and generated damage evolution data set, construct a physics-informed neural network learning model; Fuse the internal damage physical model with the physics-informed neural network learning model; Determine the physical constraint loss function to optimize the model parameters, and train the physics-informed neural network learning model to determine the hybrid model. During the training process, the weights and biases of the physics-informed neural network learning model are adaptively adjusted by minimizing the comprehensive loss function.

8. A prediction system for the internal damage process of foamed lightweight soil, the system comprising a processor, characterized in that, It also includes: A construction module for constructing an internal damage physical model of the lightweight foamed soil based on the peridynamics theory and executing the definition of the force-displacement relationship and damage evolution criterion of the non-local bond; A fusion module for fusing the internal damage physical model with the machine learning network to determine the hybrid model; An execution module for obtaining the state parameters of the lightweight foamed soil and executing the prediction of the spatio-temporal evolution of the damage field based on the hybrid model to realize the prediction of the internal damage process of the lightweight foamed soil.

9. The system according to claim 8, wherein The construction module, the fusion module, and the execution module are all data-connected to the processor.

Images