Rock-soil body fluid-solid coupling simulation method and system

Abstract

The invention relates to the technical field of fluid-solid coupling, in particular to a rock-soil body fluid-solid coupling simulation method and system. Firstly, a fluid-solid coupling model is constructed according to a rock-soil body balance equation, a seepage continuity equation and a permeability function considering deformation influence, and the fluid-solid coupling model can realize bidirectional coupling of a stress field and a seepage field, so that the fluid-solid coupling process of the rock-soil body is more accurately simulated through the fluid-solid coupling model; in addition, the fluid-structure interaction model uses many parameters, including mechanical parameters and permeability parameters, so that the inversion efficiency is low; according to the method, an orthogonal test method is adopted to carry out sensitivity analysis on the fluid-solid coupling parameters of the rock-soil body, parameters sensitive to monitoring data are obtained, only target parameters are inverted, the number of the inverted parameters is reduced, and therefore the parameter inversion efficiency is remarkably improved; the invention further provides a model parameter inversion method based on the long and short term memory neural network.

Description

Technical Field

[0001] The present invention relates to the technical field of fluid-solid coupling, and in particular to a rock-soil fluid-solid coupling simulation method and system. Background Art

[0002] Fluid-Structure Interaction (FSI) refers to the phenomenon of interaction and mutual influence between fluids and solids. By simulating the fluid-structure interaction phenomenon, geological environmental disasters can be prevented and studied; for example, landslides are a complex geological disaster process, which involves the interaction between fluids (such as groundwater) and solids (mountain rock and soil). Whether the fluid-structure interaction simulation model can be used to accurately predict the hazards caused by landslide-induced surges is the key to studying and preventing geological disasters.

[0003] Existing fluid-solid coupling simulation schemes usually only consider the effect of seepage on deformation and ignore the effect of deformation on seepage, resulting in the inability to accurately simulate the fluid-solid coupling process of rock and soil. Summary of the invention

[0004] The main purpose of the present invention is to provide a rock-soil fluid-solid coupling simulation method and system, aiming to solve the problem that the existing fluid-solid coupling simulation scheme cannot accurately simulate the rock-soil fluid-solid coupling process.

[0005] The technical solution proposed by the present invention is: A rock-soil fluid-solid coupling simulation method is applied to a rock-soil fluid-solid coupling simulation system; the system comprises a data acquisition module and a calculation module which are communicatively connected to each other; the method comprises: The calculation module establishes a fluid-solid coupling model according to the rock and soil balance equation, the seepage continuity equation and the permeability function considering the influence of deformation; The calculation module uses an orthogonal test method to perform sensitivity analysis on the parameters of the fluid-solid coupling model to obtain parameters of the fluid-solid coupling model that are sensitive to monitoring data and mark them as target parameters, wherein the parameters of the fluid-solid coupling model include mechanical parameters and permeability parameters; The calculation module establishes an objective function including displacement and groundwater level, and constructs a parameter inversion method based on a long short-term memory neural network model; The data acquisition module acquires geological data of the monitoring points in the area to be simulated and sends the data to the calculation module, wherein the geological data includes the measured displacement values ​​and the measured groundwater level values ​​corresponding to the monitoring points; The calculation module dynamically inverts the target parameters based on the parameter inversion method according to the measured displacement value and the measured groundwater level value, and substitutes the target parameters after the dynamic inversion into the fluid-solid coupling model to obtain a perfect fluid-solid coupling model; The calculation module performs fluid-solid coupling simulation on the area to be simulated based on a perfect fluid-solid coupling model.

[0006] Preferably, the calculation module establishes a fluid-solid coupling model according to the rock-soil equilibrium equation, the seepage continuity equation and the permeability function considering the influence of deformation, including: The calculation module derives the equilibrium equation of rock and soil based on Biot's three-dimensional consolidation theory: (1), In the formula, represents divergence; Indicates external force; is the elastoplastic stiffness matrix; Indicates strain; is saturation; is the pore water pressure; is the pore gas pressure; is the gradient, and in vector calculus, the gradient of a scalar field is a vector field.

[0007] Preferably, the calculation module derives the rock and soil equilibrium equation based on Biot's three-dimensional consolidation theory, and then further includes: The calculation module derives the seepage continuity equation according to Darcy's law and the principle of mass conservation: (2), In the formula, is the density of water; n is the porosity; Z is the position head; k is the permeability coefficient; div () represents the divergence operation in vector analysis, represents the derivative with respect to time; The calculation module establishes a permeability function taking deformation into account as a coupling bridge between the stress field and the seepage field, wherein the permeability function is: (3), In the formula, s is matrix suction; e is the porosity ratio, where the permeability coefficient is a function of the porosity ratio; The calculation module combines formula (1), formula (2) and formula (3) to construct a fluid-solid coupling model that can describe the bidirectional coupling between the stress field and the seepage field.

[0008] Preferably, the calculation module uses an orthogonal test method to perform sensitivity analysis on the parameters of the fluid-solid coupling model to obtain parameters of the fluid-solid coupling model that are sensitive to the monitoring data and mark them as target parameters, including: The calculation module designs a parameter combination of the fluid-solid coupling model; The calculation module takes values ​​of parameters of the fluid-solid coupling model at different levels; The calculation module performs fluid-solid coupling numerical calculation of rock and soil based on finite element simulation analysis software, inputs different numerical parameter combinations into the finite element simulation analysis software to calculate the displacement corresponding to the monitoring point in the area to be simulated, and forms a data set, wherein the data set includes each parameter combination and the corresponding calculated displacement.

[0009] Preferably, the calculation module performs fluid-solid coupling numerical calculation of rock and soil based on finite element simulation analysis software, inputs different numerical parameter combinations into the finite element simulation analysis software to calculate the displacement corresponding to the monitoring point of the area to be simulated, and forms a data set, and then further includes: The calculation module performs orthogonal test result analysis based on the data set based on finite element simulation analysis software to obtain the range corresponding to each parameter, and sorts the sensitivity of the parameters of the fluid-solid coupling model based on the range; The calculation module performs variance calculation on each parameter based on the range analysis result; The calculation module determines the parameters of the fluid-solid coupling model that are sensitive to the monitoring data by combining the range and variance calculation results, and marks them as target parameters.

[0010] Preferably, the calculation module establishes an objective function including displacement and groundwater level, and constructs a parameter inversion method based on a long short-term memory neural network model, including: The calculation module constructs a parameter inversion method based on a long short-term memory neural network model, wherein the working principle of the long short-term memory neural network model is as follows: The long short-term memory neural network model decomposes a cell unit into three parts: The first part determines what information needs to be discarded from the cell state at the previous time step, which is determined by the forget gate, where the cell state of the forget gate is: (4), In the formula, It is the sigmoid function of the long short-term memory neural network model. Each data needs to pass through the sigmoid function. represents the weight matrix; is the module output at the previous moment; Indicates that two matrices are concatenated into one total matrix; is the bias value of the total matrix; Represents the output of the forget gate; The second part determines what information is stored in the cell state. The second part consists of two parts: the sigmoid function and the tanh function. The sigmoid function cell state is updated as follows: (5), (6), In the formula, and is the weight matrix of the input gate; yes The corresponding bias term is, yes The corresponding bias term; and Both represent the state of the sigmoid function cell; tanh() is the hyperbolic tangent function, expressed as: (7), In the formula, is the hyperbolic tangent function, a nonlinear activation function commonly used in deep learning. and Both represent natural exponential functions; by using the old state Corresponding dot product , Used to discard the information that has been decided to be forgotten, and then add , which constitutes the cell state at the current moment : (8), The third part determines what information of the cell state to output, and then passes the current cell state through the tanh function and multiplies it by the output of the sigmoid function to get the final output: (9), (10) In the formula, represents the bias term, represents the weight matrix, is the output value, The cell state.

[0011] Preferably, the calculation module constructs a parameter inversion method based on a long short-term memory neural network model, comprising: The computing module trains the long short-term memory neural network model, including: The computing module constructs training samples and verification samples; The computing module initializes the weight matrix and bias term of the long short-term memory neural network model, and forwardly calculates according to the training sample to obtain the output value of each neuron; The calculation module calculates the error term value of each neuron according to the output value of each neuron, and performs back propagation calculation according to the error term value, wherein the back propagation calculation includes two parts: one is the back propagation along time, that is, from the current moment t Start calculating the previous moment t -1 error term; the second is to propagate the error term to the upper layer; The calculation module calculates the gradient of the weight matrix according to the error term value and performs an update operation.

[0012] Preferably, the calculation module constructs training samples and verification samples, including: The calculation module takes the target parameters involved in the inversion as basic variables and sets the parameter set to be inverted as , M is the number of target parameters involved in the inversion, T Represents vector transpose; The calculation module uses an orthogonal test method to obtain a parameter combination of target parameters involved in inversion; The calculation module inputs each parameter combination into the finite element simulation analysis software to calculate the displacement corresponding to the monitoring point to form a data set, and constructs a training sample and a verification sample according to the data set; The calculation module calculates the gradient of the weight matrix according to the error term value and performs an update operation, and then further includes: The calculation module uses training samples to train the long short-term memory neural network model, wherein the input value of the training sample is the calculated displacement of the monitoring point, and the output value of the training sample is the target parameter involved in the inversion; When the objective function value reaches 3% of the maximum accuracy or the maximum number of iterations, the calculation module stops training to obtain the optimal solution of the parameters of the weight matrix and the bias matrix in the long short-term memory neural network model, wherein the calculation module comprehensively considers the prior information of the displacement and groundwater level monitoring data, and the objective function constructed according to the least squares criterion is: (11), In the formula, is the number of monitoring points for groundwater level; is the number of monitoring points for displacement; is the calculated value of groundwater level at the ith groundwater level monitoring point; is the measured value of groundwater level at the ith groundwater level monitoring point; is the calculated displacement value of the ith displacement monitoring point; is the measured displacement value of the ith displacement monitoring point; The computing module uses verification samples to verify the trained long short-term memory neural network model.

[0013] Preferably, the calculation module dynamically inverts the target parameters based on the parameter inversion method according to the measured displacement values ​​and the measured groundwater level values, and substitutes the target parameters after the dynamic inversion into the fluid-solid coupling model to obtain a perfect fluid-solid coupling model, including: The calculation module inputs the obtained optimal solution of the parameters of the weight matrix and the bias matrix into the long short-term memory neural network model; The calculation module uses the measured displacement values ​​and the measured groundwater level values ​​of the monitoring points in the area to be simulated as input values ​​of the long short-term memory neural network model, and uses the target parameters to be inverted as output values; The calculation module inputs the output target parameters into the finite element simulation analysis software to calculate the displacement of the monitoring point at the current moment; The calculation module uses the displacement of the monitoring point at the current moment as the input value of the long short-term memory neural network model for updating, and uses the updated displacement to continue to invert the target parameters at the next moment, and so on to obtain the target parameters for completing the dynamic inversion.

[0014] The present invention also proposes a rock-soil fluid-solid coupling simulation system, which applies a rock-soil fluid-solid coupling simulation method; the system comprises a data acquisition module and a calculation module which are communicatively connected to each other.

[0015] Through the above technical solution, the following beneficial effects can be achieved: The rock-soil fluid-solid coupling simulation method proposed in the present invention can simulate the rock-soil fluid-solid coupling process more accurately. When used specifically, a fluid-solid coupling model is first constructed according to the rock-soil equilibrium equation, the seepage continuity equation and the permeability function considering the influence of deformation. The fluid-solid coupling model can realize the bidirectional coupling of the stress field and the seepage field, thereby more accurately simulating the rock-soil fluid-solid coupling process through the fluid-solid coupling model. In addition, considering that the fluid-solid coupling model uses many parameters, including mechanical parameters and permeability parameters, the inversion efficiency is low. The present invention uses an orthogonal test method to perform sensitivity analysis on the parameters of the rock-soil fluid-solid coupling, obtains parameters (i.e., target parameters) that are sensitive to monitoring data, and only inverts the target parameters, thereby reducing the number of inversion parameters, thereby significantly improving the parameter inversion efficiency. The present invention also proposes a parameter inversion method based on a long short-term memory neural network model, and dynamically inverts the target parameters of the fluid-solid coupling model in combination with long-sequence monitoring data, thereby obtaining more accurate model parameter values. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the structures shown in these drawings without paying creative work.

[0017] Figure 1 This is a flow chart of the first embodiment of a fluid-solid coupling simulation method for rock and soil mass proposed by the present invention. DETAILED DESCRIPTION

[0018] It should be understood that the specific embodiments described herein are only used to explain the present invention, and are not used to limit the present invention.

[0019] The invention provides a rock-soil fluid-solid coupling simulation method and system.

[0020] As attached Figure 1 As shown, in a first embodiment of a rock-soil fluid-solid coupling simulation method proposed by the present invention, the rock-soil fluid-solid coupling simulation method is applied to a rock-soil fluid-solid coupling simulation system; the system includes a data acquisition module and a calculation module that are communicatively connected to each other; the data acquisition module includes a water level meter for collecting groundwater level, and a total station for collecting geological displacement; this embodiment includes the following steps: Step S110: The calculation module establishes a fluid-solid coupling model according to the rock-soil equilibrium equation, the seepage continuity equation and the permeability function considering the influence of deformation.

[0021] Step S120: The calculation module uses an orthogonal test method to perform a sensitivity analysis on the parameters of the fluid-solid coupling model to obtain parameters of the fluid-solid coupling model that are sensitive to the monitoring data, and mark them as target parameters, wherein the parameters of the fluid-solid coupling model include mechanical parameters and permeability parameters.

[0022] Step S130: The calculation module establishes an objective function including displacement and groundwater level, and constructs a parameter inversion method based on a long short-term memory neural network model.

[0023] Step S140: The data acquisition module acquires geological data of the monitoring points in the area to be simulated and sends the data to the calculation module, wherein the geological data includes the measured displacement values ​​and the measured groundwater level values ​​corresponding to the monitoring points.

[0024] Step S150: the calculation module dynamically inverts the target parameters based on the parameter inversion method according to the measured displacement values ​​and the measured groundwater level values, and substitutes the dynamically inverted target parameters into the fluid-solid coupling model to obtain a perfect fluid-solid coupling model.

[0025] Step S160: the calculation module performs fluid-solid coupling simulation on the area to be simulated based on the perfect fluid-solid coupling model.

[0026] The rock-soil fluid-solid coupling simulation method proposed in the present invention can simulate the rock-soil fluid-solid coupling process more accurately. When used specifically, a fluid-solid coupling model is first constructed according to the rock-soil equilibrium equation, the seepage continuity equation and the permeability function considering the influence of deformation. The fluid-solid coupling model can realize the bidirectional coupling of the stress field and the seepage field, thereby more accurately simulating the rock-soil fluid-solid coupling process through the fluid-solid coupling model. In addition, considering that the fluid-solid coupling model uses many parameters, including mechanical parameters and permeability parameters, the inversion efficiency is low. The present invention uses an orthogonal test method to perform sensitivity analysis on the parameters of the rock-soil fluid-solid coupling, obtains parameters (i.e., target parameters) that are sensitive to monitoring data, and only inverts the target parameters, thereby reducing the number of inversion parameters, thereby significantly improving the parameter inversion efficiency. The present invention also proposes a parameter inversion method based on a long short-term memory neural network model, and dynamically inverts the target parameters of the fluid-solid coupling model in combination with long-sequence monitoring data, thereby obtaining more accurate model parameter values.

[0027] In a second embodiment of a fluid-solid coupling simulation method for rock and soil mass proposed by the present invention, based on the first embodiment, step S110 includes the following steps: Step S210: The calculation module derives the rock and soil equilibrium equation based on Biot's three-dimensional consolidation theory: (1), In the formula, represents divergence; Indicates external force; is the elastoplastic stiffness matrix; Indicates strain; is saturation; is the pore water pressure; is the pore gas pressure; is the gradient, and in vector calculus, the gradient of a scalar field is a vector field.

[0028] Specifically, this embodiment provides a specific formula for the rock and soil mass equilibrium equation.

[0029] In a third embodiment of a fluid-solid coupling simulation method for rock and soil mass proposed by the present invention, based on the second embodiment, step S210 further includes the following steps: Step S310: The calculation module derives the seepage continuity equation according to Darcy's law and the principle of mass conservation: (2), In the formula, is the density of water;n is the porosity; Z is the position head; k is the permeability coefficient; div () represents the divergence operation in vector analysis, represents the time derivative.

[0030] Step S320: To achieve bidirectional coupling between the stress field and the seepage field, the calculation module establishes a permeability function taking deformation into consideration as a coupling bridge between the stress field and the seepage field, wherein the permeability function is: (3), In the formula, s is matrix suction; e is the porosity ratio, where the permeability coefficient is a function of the porosity ratio.

[0031] Step S330: The calculation module combines formula (1), formula (2) and formula (3) to construct a fluid-solid coupling model that can describe the bidirectional coupling between the stress field and the seepage field.

[0032] Specifically, this embodiment provides a specific solution on how to establish a fluid-solid coupling model.

[0033] In a fourth embodiment of a fluid-solid coupling simulation method for rock and soil mass proposed by the present invention, based on the third embodiment, step S120 includes the following steps: Step S410: the calculation module designs a parameter combination of the fluid-structure coupling model.

[0034] Specifically, in this step, Minitab software is used to design the parameter combination of the fluid-solid coupling model.

[0035] Step S420: the calculation module takes values ​​of parameters of the fluid-structure coupling model at different levels.

[0036] Specifically, since parameter values ​​at different levels are crucial to accurately obtain the main parameters that are sensitive to the monitoring data, the mechanical and permeability parameter values ​​in the current prior art are used as a reference.

[0037] Step S430: The calculation module performs numerical calculation of fluid-solid coupling of rock and soil based on finite element simulation analysis software (such as ABAQUS software), inputs different numerical parameter combinations into the finite element simulation analysis software to calculate the displacement corresponding to the monitoring point of the area to be simulated, and forms a data set, wherein the data set includes each parameter combination and the corresponding calculated displacement.

[0038] Specifically, this embodiment provides a solution for determining the parameters (ie, target parameters) of the fluid-structure coupling model that are sensitive to monitoring data.

[0039] In a fifth embodiment of a fluid-solid coupling simulation method for rock and soil mass proposed by the present invention, based on the fourth embodiment, step S430 further includes the following steps: Step S510: the calculation module performs orthogonal test result analysis based on the data set based on finite element simulation analysis software to obtain the range corresponding to each parameter, and sorts the sensitivity of the parameters of the fluid-solid coupling model based on the range.

[0040] Specifically, in this embodiment, the parameters with a range less than 1 mm are set to be insensitive to the monitoring data, and in this embodiment, the parameters greater than or equal to 1 mm are selected as the target parameters.

[0041] Step S520: the calculation module performs variance calculation on each parameter based on the range analysis result in order to further test the significance level of each parameter; Specifically, for the determination of the significance level, the sensitivity of each parameter was determined by the P test (the P test is an important method for hypothesis testing in statistics. Its core is to determine whether the observed data supports the null hypothesis by calculating the P value). The specific standards are: P value < 0.01 is highly significant; 0.01 < P value < 0.05 is significant; P > 0.05 indicates insignificance.

[0042] Step S530: The calculation module determines the parameters of the fluid-structure coupling model that are sensitive to the monitoring data based on the range and variance calculation results, and marks them as target parameters.

[0043] Specifically, in this embodiment, the parameter whose range is greater than or equal to 1 mm and whose P value of variance is highly significant or significant is taken as the target parameter.

[0044] In a sixth embodiment of a fluid-solid coupling simulation method for rock and soil mass proposed by the present invention, based on the fifth embodiment, step S130 includes the following steps: Step S610: The calculation module constructs a parameter inversion method based on a long short-term memory neural network model, wherein the working principle of the long short-term memory neural network model is as follows: Specifically, the Long Short-Term Memory (LSTM) neural network is a special recurrent neural network that can effectively solve the long-term dependency problem existing in traditional recurrent neural networks; LSTM consists of a cell state, an input gate, a forget gate, and an output gate.

[0045] Cell state: The cell state is like a conveyor belt that transfers information across the sequence, allowing the information to be preserved for a long time in the sequence. Information can be added or removed from the cell state through the input gate and the forget gate. Forget gate: The forget gate determines which information in the cell state will be retained and which will be forgotten. It receives the current input and the hidden state of the previous moment as input, and outputs a value between 0 and 1, which is used to control the degree of information retention in the cell state. Input gate: The input gate is responsible for deciding which new information will be added to the cell state. It first calculates a candidate value, and then combines the output of the forget gate to update the cell state. Output gate: The output gate determines the final output based on the cell state and the current input. It first processes the cell state and then generates the output through an activation function.

[0046] In this embodiment, the long short-term memory neural network model decomposes a cell unit into three parts: The first part determines what information needs to be discarded from the cell state at the previous time step, which is determined by the forget gate, where the cell state of the forget gate is: (4), In the formula, It is the sigmoid function of the long short-term memory neural network model. Each data needs to pass through the sigmoid function. represents the weight matrix; is the module output at the previous moment; Indicates that two matrices are concatenated into one total matrix; is the bias value of the total matrix; Represents the output of the forget gate; The second part determines what information is stored in the cell state. The second part consists of two parts: the sigmoid function and the tanh function. The cell state of the sigmoid function (input gate) is updated as follows: (5), (6), In the formula, and is the weight matrix of the input gate; yes The corresponding bias term is, yes The corresponding bias term; and Both represent the state of the sigmoid function cell; tanh() is the hyperbolic tangent function, expressed as: (7), In the formula, is the hyperbolic tangent function, a nonlinear activation function commonly used in deep learning. and Both represent natural exponential functions; by using the old state Corresponding dot product , Used to discard the information that has been decided to be forgotten, and then add , which constitutes the cell state at the current moment : (8), The third part determines what information of the cell state to output, and then passes the current cell state through the tanh function and multiplies it by the output of the sigmoid function to get the final output: (9), (10) In the formula, represents the bias term, represents the weight matrix, is the output value, The cell state.

[0047] Specifically, this embodiment provides the structure of a long short-term memory neural network model.

[0048] In a seventh embodiment of a fluid-solid coupling simulation method for rock and soil mass proposed by the present invention, based on the sixth embodiment, step S610 includes the following steps: Step S710: The computing module trains the long short-term memory neural network model, including the following steps: Step S701: The calculation module constructs training samples and verification samples.

[0049] Step S702: The computing module initializes the weight matrix and bias term of the long short-term memory neural network model, and forwardly calculates according to the training samples to obtain the output value of each neuron.

[0050] Specifically, the output value of each neuron mentioned above includes The value of the vector.

[0051] Step S703: The calculation module calculates the error term value of each neuron according to the output value of each neuron, and performs back propagation calculation according to the error term value, wherein the back propagation calculation includes two parts: one is the back propagation along time, that is, from the current moment t Start calculating the previous moment t The second is to propagate the error term to the upper layer.

[0052] Step S704: The calculation module calculates the gradient of the weight matrix according to the error term value and performs an update operation.

[0053] Specifically, this embodiment provides a specific solution for constructing a parameter inversion method based on a long short-term memory neural network model.

[0054] In an eighth embodiment of a fluid-solid coupling simulation method for rock and soil mass proposed by the present invention, based on the seventh embodiment, step S701 includes the following steps: Step S810: The calculation module takes the target parameters involved in the inversion as basic variables and sets the parameter set to be inverted as , M is the number of target parameters involved in the inversion, T Represents the transpose of a vector.

[0055] Step S820: The calculation module uses an orthogonal test method to obtain a parameter combination of target parameters involved in inversion.

[0056] Step S830: The calculation module inputs each parameter combination into the finite element simulation analysis software to calculate the displacement corresponding to the monitoring point to form a data set (the data set includes each parameter combination and the corresponding displacement), and constructs training samples and verification samples according to the data set.

[0057] Step S704, and then further includes the following steps: Step S840: The calculation module uses training samples to train the long short-term memory neural network model, wherein the input value of the training sample is the calculated displacement of the monitoring point, and the output value of the training sample is the target parameter involved in the inversion.

[0058] Step S850: When the objective function value reaches 3% of the maximum accuracy or the maximum number of iterations, the calculation module stops training to obtain the optimal solution of the parameters of the weight matrix and the bias matrix in the long short-term memory neural network model, wherein the calculation module comprehensively considers the prior information of the displacement and groundwater level monitoring data, and the objective function constructed according to the least squares criterion is: (11), In the formula, is the number of monitoring points for groundwater level; is the number of monitoring points for displacement; is the calculated value of groundwater level at the ith groundwater level monitoring point; is the measured value of groundwater level at the ith groundwater level monitoring point; is the calculated displacement value of the ith displacement monitoring point; is the measured displacement value of the ith displacement monitoring point.

[0059] Step S860: The computing module uses a verification sample to verify the trained long short-term memory neural network model.

[0060] Specifically, the above steps provide a specific solution for how to train and verify the long short-term memory neural network model.

[0061] In a ninth embodiment of a fluid-solid coupling simulation method for rock and soil mass proposed by the present invention, based on the eighth embodiment, step S150 includes the following steps: Step S910: The calculation module inputs the obtained optimal solution of the parameters of the weight matrix and the bias matrix into the long short-term memory neural network model.

[0062] Step S920: The calculation module uses the measured displacement values ​​and the measured groundwater level values ​​of the monitoring points in the area to be simulated as input values ​​of the long short-term memory neural network model, and uses the target parameters to be inverted as output values.

[0063] Step S930: The calculation module inputs the output target parameters into the finite element simulation analysis software to calculate the displacement of the monitoring point at the current moment.

[0064] Step S940: The calculation module uses the displacement of the monitoring point at the current moment as the input value of the long short-term memory neural network model for updating, and uses the updated displacement to continue to invert the target parameters at the next moment, and so on, to obtain the target parameters for completing the dynamic inversion.

[0065] Specifically, this embodiment provides a specific solution on how to improve the fluid-solid coupling model through dynamic inversion.

[0066] The present invention also proposes a rock-soil fluid-solid coupling simulation system, which applies a rock-soil fluid-solid coupling simulation method; the system comprises a data acquisition module and a calculation module which are communicatively connected to each other.

[0067] The serial numbers of the above embodiments of the present invention are only for description and do not represent the advantages or disadvantages of the embodiments.

[0068] The embodiments of the present invention are described above in conjunction with the accompanying drawings, but the present invention is not limited to the above-mentioned specific implementation modes, which are merely illustrative rather than restrictive. Under the guidance of the present invention, ordinary technicians in this field can also make many forms without departing from the scope of protection of the present invention and the claims, all of which are within the protection of the present invention.

Claims

1. A fluid-solid coupling simulation method for rock and soil, characterized in that: Applicable to a fluid-solid coupling simulation system for rock and soil; the system comprises a data acquisition module and a calculation module which are communicatively connected to each other; the method comprises: The calculation module establishes a fluid-solid coupling model according to the rock and soil balance equation, the seepage continuity equation and the permeability function considering the influence of deformation; The calculation module uses an orthogonal test method to perform sensitivity analysis on the parameters of the fluid-solid coupling model to obtain parameters of the fluid-solid coupling model that are sensitive to monitoring data and mark them as target parameters, wherein the parameters of the fluid-solid coupling model include mechanical parameters and permeability parameters; The calculation module establishes an objective function including displacement and groundwater level, and constructs a parameter inversion method based on a long short-term memory neural network model; The data acquisition module acquires geological data of the monitoring points in the area to be simulated and sends the data to the calculation module, wherein the geological data includes the measured displacement values ​​and the measured groundwater level values ​​corresponding to the monitoring points; The calculation module dynamically inverts the target parameters based on the parameter inversion method according to the measured displacement value and the measured groundwater level value, and substitutes the target parameters after the dynamic inversion into the fluid-solid coupling model to obtain a perfect fluid-solid coupling model; The calculation module performs fluid-solid coupling simulation on the area to be simulated based on a perfect fluid-solid coupling model.

2. A rock-soil fluid-solid coupling simulation method according to claim 1, characterized in that: The calculation module establishes a fluid-solid coupling model according to the rock and soil equilibrium equation, the seepage continuity equation and the permeability function considering the influence of deformation, including: The calculation module derives the equilibrium equation of rock and soil based on Biot's three-dimensional consolidation theory: （1）， In the formula, represents divergence; Indicates external force; is the elastoplastic stiffness matrix; Indicates strain; is saturation; is the pore water pressure; is the pore gas pressure; is the gradient, and in vector calculus, the gradient of a scalar field is a vector field.

3. A rock-soil fluid-solid coupling simulation method according to claim 2, characterized in that: The calculation module derives the equilibrium equation of rock and soil based on Biot's three-dimensional consolidation theory, and then includes: The calculation module derives the seepage continuity equation according to Darcy's law and the principle of mass conservation: （2）， In the formula, is the density of water; n is the porosity; Z is the position head; k is the permeability coefficient; div () represents the divergence operation in vector analysis, represents the derivative with respect to time; The calculation module establishes a permeability function taking deformation into account as a coupling bridge between the stress field and the seepage field, wherein the permeability function is: （3）， In the formula, s is matrix suction; e is the porosity ratio, where the permeability coefficient is a function of the porosity ratio; The calculation module combines formula (1), formula (2) and formula (3) to construct a fluid-solid coupling model that can describe the bidirectional coupling between the stress field and the seepage field.

4. A rock-soil fluid-solid coupling simulation method according to claim 3, characterized in that: The calculation module uses an orthogonal test method to perform sensitivity analysis on the parameters of the fluid-solid coupling model to obtain parameters of the fluid-solid coupling model that are sensitive to the monitoring data and mark them as target parameters, including: The calculation module designs a parameter combination of the fluid-solid coupling model; The calculation module takes values ​​of parameters of the fluid-solid coupling model at different levels; The calculation module performs fluid-solid coupling numerical calculation of rock and soil based on finite element simulation analysis software, inputs different numerical parameter combinations into the finite element simulation analysis software to calculate the displacement corresponding to the monitoring point in the area to be simulated, and forms a data set, wherein the data set includes each parameter combination and the corresponding calculated displacement.

5. A rock-soil fluid-solid coupling simulation method according to claim 4, characterized in that: The calculation module performs fluid-solid coupling numerical calculation of rock and soil based on finite element simulation analysis software, inputs different numerical parameter combinations into the finite element simulation analysis software to calculate the displacement corresponding to the monitoring point of the area to be simulated, and forms a data set, and then also includes: The calculation module performs orthogonal test result analysis based on the data set based on finite element simulation analysis software to obtain the range corresponding to each parameter, and sorts the sensitivity of the parameters of the fluid-solid coupling model based on the range; The calculation module performs variance calculation on each parameter based on the range analysis result; The calculation module determines the parameters of the fluid-solid coupling model that are sensitive to the monitoring data by combining the range and variance calculation results, and marks them as target parameters.

6. A rock-soil fluid-solid coupling simulation method according to claim 5, characterized in that: The calculation module establishes an objective function including displacement and groundwater level, and constructs a parameter inversion method based on a long short-term memory neural network model, including: The calculation module constructs a parameter inversion method based on a long short-term memory neural network model, wherein the working principle of the long short-term memory neural network model is as follows: The long short-term memory neural network model decomposes a cell unit into three parts: The first part determines what information needs to be discarded from the cell state at the previous time step, which is determined by the forget gate, where the cell state of the forget gate is: （4）， In the formula, It is the sigmoid function of the long short-term memory neural network model. Each data needs to pass through the sigmoid function. represents the weight matrix; is the module output at the previous moment; Indicates that two matrices are concatenated into one total matrix; is the bias value of the total matrix; Represents the output of the forget gate; The second part determines what information is stored in the cell state. The second part consists of two parts: the sigmoid function and the tanh function. The sigmoid function cell state is updated as follows: （5）， （6）， In the formula, and is the weight matrix of the input gate; yes The corresponding bias term is, yes The corresponding bias term; and Both represent the state of the sigmoid function cell; tanh() is the hyperbolic tangent function, expressed as: （7）， In the formula, is the hyperbolic tangent function, and Both represent natural exponential functions; by using the old state Corresponding dot product , Used to discard the information that has been decided to be forgotten, and then add , which constitutes the cell state at the current moment : （8）， The third part determines what information of the cell state to output, and then passes the current cell state through the tanh function and multiplies it by the output of the sigmoid function to get the final output: （9）， （10） In the formula, represents the bias term, represents the weight matrix, is the output value, The cell state.

7. A rock-soil fluid-solid coupling simulation method according to claim 6, characterized in that: The calculation module constructs a parameter inversion method based on a long short-term memory neural network model, including: The computing module trains the long short-term memory neural network model, including: The computing module constructs training samples and verification samples; The computing module initializes the weight matrix and bias term of the long short-term memory neural network model, and forwardly calculates according to the training sample to obtain the output value of each neuron; The calculation module calculates the error term value of each neuron according to the output value of each neuron, and performs back propagation calculation according to the error term value, wherein the back propagation calculation includes two parts: one is the back propagation along time, that is, from the current moment t Start calculating the previous moment t -1 error term; the second is to propagate the error term to the upper layer; The calculation module calculates the gradient of the weight matrix according to the error term value and performs an update operation.

8. A rock-soil fluid-solid coupling simulation method according to claim 7, characterized in that: The calculation module constructs training samples and verification samples, including: The calculation module takes the target parameters involved in the inversion as basic variables and sets the parameter set to be inverted as , M is the number of target parameters involved in the inversion, T Represents vector transpose; The calculation module uses an orthogonal test method to obtain a parameter combination of target parameters involved in inversion; The calculation module inputs each parameter combination into the finite element simulation analysis software to calculate the displacement corresponding to the monitoring point to form a data set, and constructs a training sample and a verification sample according to the data set; The calculation module calculates the gradient of the weight matrix according to the error term value and performs an update operation, and then further includes: The calculation module uses training samples to train the long short-term memory neural network model, wherein the input value of the training sample is the calculated displacement of the monitoring point, and the output value of the training sample is the target parameter involved in the inversion; When the objective function value reaches 3% of the maximum accuracy or the maximum number of iterations, the calculation module stops training to obtain the optimal solution of the parameters of the weight matrix and the bias matrix in the long short-term memory neural network model, wherein the calculation module comprehensively considers the prior information of the displacement and groundwater level monitoring data, and the objective function constructed according to the least squares criterion is: （11）， In the formula, is the number of monitoring points for groundwater level; is the number of monitoring points for displacement; is the calculated value of groundwater level at the ith groundwater level monitoring point; is the measured value of groundwater level at the ith groundwater level monitoring point; is the calculated displacement value of the ith displacement monitoring point; is the measured displacement value of the ith displacement monitoring point; The computing module uses verification samples to verify the trained long short-term memory neural network model.

9. A rock-soil fluid-solid coupling simulation method according to claim 8, characterized in that: The calculation module dynamically inverts the target parameters based on the parameter inversion method according to the measured displacement value and the measured groundwater level value, and substitutes the target parameters after the dynamic inversion into the fluid-solid coupling model to obtain a perfect fluid-solid coupling model, including: The calculation module inputs the obtained optimal solution of the parameters of the weight matrix and the bias matrix into the long short-term memory neural network model; The calculation module uses the measured displacement values ​​and the measured groundwater level values ​​of the monitoring points in the area to be simulated as input values ​​of the long short-term memory neural network model, and uses the target parameters to be inverted as output values; The calculation module inputs the output target parameters into the finite element simulation analysis software to calculate the displacement of the monitoring point at the current moment; The calculation module uses the displacement of the monitoring point at the current moment as the input value of the long short-term memory neural network model for updating, and uses the updated displacement to continue to invert the target parameters at the next moment, and so on to obtain the target parameters for completing the dynamic inversion.

10. A rock-soil fluid-solid coupling simulation system, characterized in that: The fluid-solid coupling simulation method for rock and soil mass as described in any one of claims 1 to 9 is applied; the system comprises a data acquisition module and a calculation module which are communicatively connected to each other.

Images