Method and system for predicting stability of reverse-wrapping geotextile slope

Abstract

The invention provides a stability prediction method and system for an anti-package geotextile slope, and relates to the technical field of stability prediction.The method comprises the steps that design data and historical monitoring data of the anti-package geotextile slope are obtained, and a mathematical optimization model is built based on the design data and the historical monitoring data; stability calculation is carried out based on the mathematical optimization model, a data set is generated based on safety coefficient data, design data and historical monitoring data obtained through calculation, and the data set comprises a training set and a test set; training, optimizing and testing a preset neural network based on a preset improved dung beetle optimization algorithm and the training set to obtain a tested slope safety coefficient prediction model; and preset real-time monitoring data are sent to the slope safety coefficient prediction model for prediction, and the stability value of the anti-wrapping geotextile slope is obtained. According to the method, various parameters can be considered, the adaptability, the fault-tolerant capability and the self-perfection capability are very high, and the monitoring and calculation accuracy is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of stability prediction, and more specifically, to a method and system for predicting the stability of an anti-wrapped geotextile slope. Background Art

[0002] Currently, during the later maintenance and prevention of an anti-wrapped geotextile slope, manual analysis, monitoring, and calculation of the slope's stress, stability, etc. are still carried out. It is greatly affected by human factors. Considering the number of parameters and the mutual coupling of various parameters increases the manual time cost and monitoring difficulty, and increases the failure probability of the structure. Therefore, in order to prevent disasters, a large number of manual operations are carried out for later monitoring, resulting in a great waste of time cost and economic cost, and invisibly increasing the construction cost.

[0003] Moreover, there are numerous hyperparameters in traditional calculation models, which are difficult to select manually, and the training process is very slow, having a great impact on performance loss. Therefore, there is a need for a method and system for predicting the stability of an anti-wrapped geotextile slope that can improve the optimization performance and operation speed, so as to improve the accuracy of predicting the stability of the anti-wrapped geotextile slope and reduce costs. Summary of the Invention

[0004] The purpose of the present invention is to provide a method and system for predicting the stability of an anti-wrapped geotextile slope to improve the above problems. To achieve the above purpose, the technical solutions adopted by the present invention are as follows:

[0005] In a first aspect, the present application provides a method for predicting the stability of an anti-wrapped geotextile slope, including:

[0006] Obtain the design data and historical monitoring data of the anti-wrapped geotextile slope, and build a FreeMat mathematical optimization model based on the design data and historical monitoring data;

[0007] Input the design data and historical monitoring data of the anti-wrapped geotextile slope into the FreeMat mathematical optimization model for stability calculation, and generate a data set based on the calculated safety factor data, design data, and historical monitoring data. The data set includes a training set and a test set;

[0008] Train and optimize a preset neural network based on a preset improved dung beetle optimization algorithm and the training set to obtain a slope safety factor prediction model;

[0009] Test the stability prediction model based on the test set to obtain a slope safety factor prediction model after testing;

[0010] Send the preset real-time monitoring data to the slope safety factor prediction model for prediction to obtain the safety factor corresponding to each real-time monitoring data, and use the safety factor corresponding to each real-time monitoring data as the stability value of the wrapped geotextile slope.

[0011] In a second aspect, the present application also provides a stability prediction system for a wrapped geotextile slope, including:

[0012] An acquisition unit, configured to acquire the design data and historical monitoring data of the wrapped geotextile slope, and build a FreeMat mathematical optimization model based on the design data and historical monitoring data;

[0013] A calculation unit, configured to input the design data and historical monitoring data of the wrapped geotextile slope into the FreeMat mathematical optimization model for stability calculation, and generate a data set based on the calculated safety factor data, design data, and historical monitoring data, where the data set includes a training set and a test set;

[0014] A training unit, configured to train and optimize a preset neural network based on a preset improved dung beetle optimization algorithm and the training set to obtain a slope safety factor prediction model;

[0015] A test unit, configured to test the stability prediction model based on the test set to obtain a slope safety factor prediction model after testing;

[0016] A prediction unit, configured to send the preset real-time monitoring data to the slope safety factor prediction model for prediction to obtain the safety factor corresponding to each real-time monitoring data, and use the safety factor corresponding to each real-time monitoring data as the stability value of the wrapped geotextile slope.

[0017] The beneficial effects of the present invention are as follows:

[0018] The improved stability prediction system for the wrapped geotextile slope in the present invention has a relatively high operation speed, can reduce the time cost, and at the same time reduce the performance loss. The stability prediction system for the wrapped geotextile slope in the present invention can consider various parameters, has strong adaptability, fault tolerance ability, and self-improving ability, and improves the accuracy of monitoring and calculation; the stability prediction system for the wrapped geotextile slope in the present invention can automatically search for solutions that meet any requirements according to user needs, reducing the labor cost of monitoring and calculation.

[0019] Other features and advantages of the present invention will be described in the subsequent description, and part of them will become obvious from the description, or will be understood by implementing the embodiments of the present invention. The objectives and other advantages of the present invention can be achieved and obtained through the structures specifically pointed out in the written description, claims, and drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following will briefly introduce the drawings required for use in the embodiments. It should be understood that the following drawings only show certain embodiments of the present invention, and therefore should not be regarded as limiting the scope. For those of ordinary skill in the art, without creative efforts, other related drawings can also be obtained based on these drawings.

[0021] Figure 1 It is a schematic flowchart of the method for predicting the stability of the wrapped geotextile slope described in the embodiments of the present invention;

[0022] Figure 2 It is a schematic structural diagram of the system for predicting the stability of the wrapped geotextile slope described in the embodiments of the present invention.

[0023] In the figure: 701, acquisition unit; 702, calculation unit; 703, training unit; 704, testing unit; 705, prediction unit. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0024] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the drawings in the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. Usually, the components of the embodiments of the present invention described and shown in the drawings here can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the drawings is not intended to limit the scope of the claimed present invention, but merely represents the selected embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts belong to the scope of protection of the present invention.

[0025] It should be noted that: similar reference numerals and letters indicate similar items in the following drawings. Therefore, once an item is defined in one drawing, it does not need to be further defined and explained in subsequent drawings. At the same time, in the description of the present invention, the terms "first", "second", etc. are only used for distinguishing descriptions and cannot be understood as indicating or implying relative importance.

[0026] Embodiment 1:

[0027] This embodiment provides a method for predicting the stability of a wrapped geotextile slope.

[0028] See Figure 1 , the figure shows that this method includes steps S1, S2, S3, S4, and S5.

[0029] Step S1: Obtain the design data and historical monitoring data of the reversed-encased geotextile slope, and build a FreeMat mathematical optimization model based on the design data and historical monitoring data;

[0030] It can be understood that the design data of the reversed-encased geotextile slope includes the material physical and mechanical parameters (unit weight γ, cohesion c, internal friction angle ) of each soil layer of the slope rock and soil mass, the slope surface equation e 0 (x) of the slope design, the dividing line equation e j (x) of each soil layer, and the slip line equation y k (x) of each soil layer; the historical monitoring data of the reversed-encased geotextile slope includes the slip surface parameters (the abscissa of the preset control point, the left and right intersection points x s , x e ) of each preset control point; in this step, the design data and historical monitoring data of the reversed-encased geotextile slope are also input into the FreeMat programming software to build a mathematical optimization model, and then the FreeMat mathematical optimization model is obtained.

[0031] Step S2: Input the design data and historical monitoring data of the reversed-encased geotextile slope into the FreeMat mathematical optimization model for stability calculation, and generate a data set based on the calculated safety factor data, design data, and historical monitoring data. The data set includes a training set and a test set;

[0032] It can be understood that in this step, Step S2 includes Step S21, Step S22, Step S23, and Step S24.

[0033] Step S21: Determine the slope slice width and the number of control points based on the design data and historical monitoring data, and construct an abscissa matrix based on the slope slice width and the number of control points;

[0034] It can be understood that in this step, the slope slice width o is calculated, and the abscissas of n + 1 control points are obtained in sequence and stored in the matrix X:

[0035] X = [x s , x s + o,..., x s + o i × (i - 1),..., x e , i = 1, 2,..., n + 1

[0036] o = (x e - x s ) / n

[0037] In the formula, x s , x eRepresent the left and right intersection points of the sliding surface and the slope surface; o represents the strip width; o i The strip width of the i-th strip.

[0038] Step S22: Calculate the ordinate value of each control point based on the abscissa matrix and the preset sliding surface equation;

[0039] It can be understood that in this step, according to the matrix X, the ordinate Y of n + 1 control points is obtained by substituting into the sliding surface equation;

[0040] Specifically, the sliding surface equation and the ordinate Y of n + 1 control points are calculated as follows;

[0041]

[0042] In the formula, (x o , y o ) represents the center coordinates; r represents the distance between the center coordinates and any point on the arc; x s , x e represent the left and right intersection points of the sliding surface and the slope surface.

[0043] Step S23: Calculate the bottom slope angles of a preset number of strips based on the abscissa matrix and the ordinate value of each control point to obtain the bottom slope angle matrix;

[0044] It can be understood that in this step, the bottom slope angles a i of n strips are calculated in sequence according to the ordinate Y and stored in the row matrix a:

[0045] Specifically, the bottom slope angles a i of n strips are calculated as follows;

[0046]

[0047] In the formula, atand() is the inverse function of the tangent function y = tanx in the FreeMat software; X(i) and Y(i) respectively represent the abscissa and ordinate matrices of the control points of the i-th strip.

[0048] Step S24: Calculate the safety factor value of the slope based on the design data, historical monitoring data, abscissa matrix, ordinate value of each control point, and bottom slope angle matrix.

[0049] It can be understood that in this step, the physical and mechanical parameters (unit weight γ, cohesion c, internal friction angle ) of each soil layer material are captured according to the control point coordinate information; the physical and mechanical parameter values of each soil layer material are respectively stored in the vectors c, and γ

[0050]

[0051] where γ i represents the unit weight of the i-th slice, c i represents the cohesion of the i-th slice, and φ represents the internal friction angle of the i-th slice;

[0052] It can be understood that in this step, according to the physical and mechanical parameters of the soil layer material, it is assumed that the dividing line equations of each soil layer from top to bottom are e 1 (x), e 2 (x),..., e i (x),..., e n (x), and the weight W i of slice i is calculated;

[0053] Specifically, the weight W i of slice i is calculated as follows:

[0054] If e 1 (X(i)) ≤ Y(i) ≤ e 0 (X(i)), and e 1 (X(i + 1)) ≤ Y(i + 1) ≤ e 0 (X(i + 1)), then the slip line at the bottom of slice i is in the first soil layer; it is easy to obtain the material parameter values of soil layer 1 as c = c(1), γ = γ(1), then the weight W i of slice i is:

[0055]

[0056] where W i represents the weight of slice i; h i represents the average abscissa of slice i and slice i + 1; γ i represents the unit weight of the soil layer of the i-th slice; X(i), X(i + 1) represent the abscissas of slice i and slice i + 1; e o (x) represents the slope surface equation of the slope design, and e o (h i ) is the function value corresponding to the abscissa h i ; y(x) represents the slip surface equation of the slope design, and y(h i ) is the function value corresponding to the abscissa h i ;

[0057] If e j (X(i)) ≤ Y(i) ≤ e j-1 (X(i)), and e j (X(i + 1)) ≤ Y(i + 1) ≤ e j-1(X(i + 1)), (j = 2, ..., n), then the slip line at the bottom of slice i is in the j-th layer of soil; thus, the material parameter value c of soil layer 1 can be obtained j = c(j), γ j = γ(j), then the gravity W of slice i i is:

[0058]

[0059] In the formula, W i represents the gravity of slice i; h i represents the average abscissa of slice i and slice i + 1; γ j represents the unit weight of the j-th slice of soil layer; X(i), X(i + 1) represent the abscissas of slice i and slice i + 1; e j (x) represents the dividing line equation of each soil layer, e j (h i ) is the function value corresponding to the abscissa h i ; y(x) represents the slip surface equation of the slope design, y(h i ) is the function value corresponding to the abscissa h i ;

[0060] It can be understood that in this step, according to the gravity W of the slice i i and the physical and mechanical parameters of the soil layer material, the safety factor of the slope is calculated.

[0061] Specifically, the safety factor of the slope is calculated as follows:

[0062]

[0063] In the formula, F s represents the safety factor of the slope; W i represents the gravity of slice i; h i represents the average abscissa of slice i and slice i + 1; represents the internal friction angle of the j-th slice of soil layer; c j represents the cohesion of the j-th slice of soil layer; X(i), X(i + 1) represent the abscissas of slice i and slice i + 1; a(i) represents the bottom inclination angle of the i-th slice; n ai is an intermediate variable.

[0064] It can be understood that in this step, the design data and historical monitoring data of the reverse-wrapped geotextile slope are also imported into finite difference software to generate a random field image. Multiple data samples are generated from the random field image, and the data samples are brought into the mathematical optimization model to calculate the slope safety factor corresponding to each data sample. A data set is composed of all the data samples, and the data set is divided into a training set and a test set according to a ratio of 7:3;

[0065] Step S3: Train and optimize a preset neural network based on a preset improved dung beetle optimization algorithm and the training set to obtain a slope safety factor prediction model;

[0066] It can be understood that in this step, the neural network is optimized by the improved dung beetle optimization algorithm, which is an improvement based on the dung beetle optimization algorithm. When the dung beetle algorithm solves optimization problems, it usually initializes the population randomly. However, this random initialization may cause the algorithm to be too concentrated in certain areas within the search space, thus falling into a local optimum. By introducing Tent chaotic mapping during the population initialization process, the diversity of the initial population can be effectively improved, which helps the algorithm to conduct a more extensive global search in the early stage, thereby accelerating the convergence speed and improving the solution accuracy. Secondly, average differential mutation perturbation, Levy-Cauchy mutation, and greedy selection strategy are used for screening. The fitness values of the screened excellent dung beetle populations are calculated, and the dung beetle populations are updated based on the calculated fitness values until the number of iterations reaches the maximum number of iterations, obtaining the optimal dung beetle population information and the corresponding hidden layer parameters of the neural network. The safety factor of the most accurate slope is output through the optimal hidden layer parameters of the neural network to judge the stability of the slope. The more accurate prediction of the slope stability is realized. The optimized model can better utilize complex and diverse slope data information, improve the robustness and accuracy of the prediction model, and provide important technical support for engineering decisions. In this step, step S3 includes step S31, step S32, step S33, step S34, and step S35.

[0067] Step S31: Build a bidirectional recurrent neural network model based on an input layer, a forward hidden layer, a backward hidden layer, a fully connected layer, and an output layer;

[0068] It can be understood that the bidirectional recurrent neural network model in this step includes an input layer, a forward hidden layer, a backward hidden layer, a fully connected layer, and an output layer.

[0069] Step S32: Initialize the weights and parameters of the bidirectional recurrent neural network unit using a standard normal distribution to obtain the initialization parameters;

[0070] It can be understood that based on the pytorch framework, the weights and parameters of the fully connected layer are initialized using a standard normal distribution.

[0071] Step S33: Perform temporal modeling based on the bidirectional recurrent neural network unit and the hidden layer, and output the hidden state at each moment;

[0072] It can be understood that in this step, temporal modeling is performed based on the bidirectional recurrent neural network unit and the hidden layer, and the hidden state h at time t is output. t ;

[0073] Specifically, the hidden state h at time t t is calculated as follows

[0074]

[0075] In the formula, h t represents the hidden state at time t; h t-1 represents the hidden state at the previous moment; x t represents the input at the current moment; Z t represents the update gate of the bidirectional recurrent neural network unit; r t represents the reset gate of the bidirectional recurrent neural network unit; W, U, and b are the weights and biases of the update gate or the reset gate; represents the candidate hidden state at time t, which is used to update the hidden state; tanh is the activation function; Sigmoid is the S-shaped function;

[0076] It can be understood that the bidirectional recurrent neural network model consists of recurrent neural networks in two directions (forward and backward), and the basic unit of the recurrent neural network in each direction is the gated recurrent unit. The gated recurrent unit consists of an update gate (Z t ) and a reset gate (r t ). The update gate defines the amount of the previous memory saved to the current moment, and the reset gate determines the influence degree of the past hidden state at the current moment. The value ranges of both are between 0 and 1.

[0077] Step S34: Use the hidden state parameters of the reverse hidden layer in the bidirectional recurrent neural network model as hyperparameters, and optimize the hyperparameters based on the preset improved dung beetle optimization algorithm to obtain the optimized hyperparameters;

[0078] It can be understood that in this step, the hidden state parameters are used as hyperparameters. According to the initial range of the hyperparameters and set the parameters of the preset improved dung beetle optimization algorithm. The parameters include the deflection coefficient, constant b, problem dimension, maximum number of iterations, upper and lower bounds of the optimization problem, and random numbers subject to normal distribution. It can be understood that in this step, step S34 includes step S341, step S342, step S343, and step S344.

[0079] Step S341: Initialize all the hyperparameters as input data to obtain an initialized dung beetle population;

[0080] It can be understood that in this step, a dung beetle population with a population size of m is randomly generated. By introducing the Tent chaos mapping method, the dung beetle population is evenly distributed and the initial positions of the dung beetle population are changed, and a better distribution of the dung beetle population can be obtained.

[0081] Specifically, the Tent chaos mapping is described as follows:

[0082]

[0083] where: β is the chaos parameter, and its range of variation is (0, 1]; Z n is the position of the dung beetle population corresponding to the nth iteration of the Tent chaos mapping; Z n+1 is the position of the dung beetle population corresponding to the (n + 1)th iteration of the Tent chaos mapping.

[0084] Step S342: Screen the dung beetle population by using average differential mutation perturbation, Levy-Cauchy mutation, and greedy selection strategy to obtain the screened dung beetle population;

[0085] It can be understood that in this step, the dung beetle population is screened by using average differential mutation perturbation, Levy-Cauchy mutation, and greedy selection strategy to obtain the screened dung beetle population. The specific descriptions of its three main stages are as follows:

[0086] Step1: Average differential mutation perturbation

[0087]

[0088] In the formula: X b represents the current best dung beetle population; X i represents the individual that is currently mutating; F is the scaling factor; X r1 and X r2 represent two randomly selected dung beetle populations from the current population; X c1 and X c2 represent two new dung beetle populations updated according to two randomly selected dung beetle populations from the current population;

[0089] It can be understood that in order to prevent the algorithm from converging prematurely due to the reduction of population diversity during the entire iteration process, average differential mutation is introduced. According to different iteration stages, this method can be divided into two variants. Both of these two variants initially randomly select two dung beetle populations X r1 and X r2 from the current population, and calculate two new dung beetle populations X c1 and X c2。The uniqueness of the first mutation strategy lies in that it adopts two basic vectors outside the current population. This strategy not only helps to avoid the problem of population stagnation, but also effectively maintains the diversity of the population, thus enhancing the exploration ability of the algorithm. Therefore, the algorithm can search in a wider solution space, increasing the possibility of finding the global optimal solution. The second mutation strategy is such that the generation of the new population contains information about the global optimal solution. This improvement allows the algorithm to perform a more intensive search near the optimal solution, thus finely exploring small changes in the solution space. In this way, the algorithm can approximate the global optimal solution more precisely, improving the accuracy and efficiency of the solution.

[0090] Step2: Levy-Cauchy Mutation

[0091] After the previous stage is completed, by performing Levy-Cauchy mutation on the perturbation parameter vector, the perturbation parameters are made diverse. For this purpose, a composite adaptive mutation strategy combining Levy mutation and Cauchy mutation is designed. This strategy is applied to the mutation process of the global optimal position, significantly enhancing the ability of the algorithm to escape from local optimal solutions and at the same time broadening the search scope of the algorithm. Its mathematical expression is as follows:

[0092]

[0093] In the formula, Y b (t) represents the position of the optimal position X b (t) of the dung beetle population after being perturbed by Levy-Cauchy mutation; X b (t) is the optimal position of the dung beetle population in the t-th iteration process; t is the current iteration number, T is the maximum iteration number, and σ is the mutation factor.

[0094] Step3: Greedy Selection Strategy

[0095] After the initial dung beetle population undergoes the average differential mutation perturbation and the Levy-Cauchy mutation stage, the trial vector U i (t + 1) is compared with the target vector X i (t) using the greedy principle to determine whether it can become an iterative individual of the next generation of the dung beetle population. If the fitness value of the experimental vector U i (t + 1) is less than the target vector X i (t), then X i+1 (t + 1) is set to U i (t + 1); otherwise, the old value X i (t) is retained.

[0096] Its expression description is:

[0097]

[0098] wherein, U i (t + 1) represents the experimental vector; X i (t) represents the target vector.

[0099] Step S343: Calculate the fitness value of the screened dung beetle population based on a preset fitness function, and update the position of the dung beetle population based on the fitness value until the number of iterations reaches the maximum number of iterations, so as to obtain the optimal dung beetle population as the position information;

[0100] It can be understood that the preset fitness function in this step is as follows:

[0101]

[0102] wherein, x i (t) represents the position of the i-th dung beetle population at the t-th iteration; y i represents the output result of the neural network model, and Q i represents the expected value.

[0103] After calculating the fitness value, judge whether the fitness value is less than a preset fitness threshold. If it is less, execute the following steps;

[0104] Step 1: Dung beetle ball-rolling stage. Set a preset judgment threshold of 0.9. When δ < 0.9, update the dung beetle individual, and the formula is as follows:

[0105] x i (t + 1) = x i (t) + α·k·x i (t - 1) + b·Δx

[0106] When δ ≥ 0.9, update the position of the dung beetle individual, and the formula is as follows:

[0107] x i (t + 1) = x i (t) + tanθ·|x i (t) - x i (t - 1)|

[0108] In the formula, x i (t) represents the position of the i-th dung beetle at the t-th generation; t represents the current number of iterations; x i (t + 1) represents the position of the i-th dung beetle after update at the (t + 1)-th generation; α is a natural coefficient with a value of -1 or 1. When α = -1, it means that the dung beetle individual deviates from the original ball-rolling direction, and when α = 1, it means there is no deviation; the parameter k ∈ (0, 0.2] represents the deflection coefficient, the parameter b ∈ (0, 1]; θ ∈ (0, π] represents the deflection angle. When θ takes the values of 0, π / 2, or π, the position of the dung beetle individual will not be updated; Δx = |xi (t)-X ω |, X ω represents the global worst position of the current dung beetle individual;

[0109] Step 2: Egg-laying stage. If t ≤ 3, update the position of the egg ball individual. The formula is as follows:

[0110] B i (t + 1) = X * + b 1 × (B i (t) - Lb * ) + b 2 × (B i (t) - Ub * )

[0111] If t > 3, then store the boundary information of the egg-laying areas of the previous t - 1, t - 2, and t - 3 generations, calculate the evolutionary factor f to correct the fractional order v, adjust the boundaries of the new generation, and update the position of the egg ball individual based on the boundaries of the new generation;

[0112] The dynamic lower and upper boundary adjustment rules can be expressed as:

[0113]

[0114] In the formula, B i (t) represents the position of the i-th egg ball at generation t; X * represents the local best position of the current dung beetle individual; Lb* and Ub* respectively represent the lower and upper boundaries of the egg-laying area of the female dung beetle; b 1 and b 2 are two random numbers belonging to (0, 1); f represents the evolutionary factor; υ represents the fractional order;

[0115] Step 3: Food-foraging stage of the young dung beetles. If t ≤ 3, update the position of the young dung beetle individual. The formula is as follows:

[0116] X i (t + 1) = X i (t) + C 1 × (X i (t) - Lb b ) + C 2 × (X i (t) - Ub b )

[0117] If t > 3, then store the boundary information of the food-foraging areas of the young dung beetles of the previous t - 1, t - 2, and t - 3 generations, calculate the evolutionary factor f to correct the fractional order v, adjust the boundaries of the new generation, and update the position of the young dung beetle individual based on the boundaries of the new generation; The dynamic lower and upper boundary adjustment rules can be expressed as:

[0118]

[0119] where X i (t) represents the position of the i-th individual dung beetle at the t-th generation; C 1 and C 2 are two random numbers belonging to (0, 1); f represents the evolution factor; υ represents the fractional order; Lb b and Ub b represent the lower and upper boundaries of the foraging area of the dung beetle, respectively;

[0120] Step 4: Dung beetle stealing stage, update the position of the exploring dung beetle, and detect the boundary information of the exploring dung beetle to prevent it from exceeding the boundary. The formula is as follows:

[0121] X e (t + 1) = X e (t) + 2r(X e (t) - X e (t - 1)) + A

[0122] where X e (t - 1), X e (t), X e (t + 1) represent the positions of the exploring dung beetles at generations t - 1, t, and t + 1, respectively; r is the moving step factor of the exploring dung beetle, which is a random number within [0, 1]; A is the moving step of the exploring dung beetle;

[0123] Step 5: Calculate the average position X mean of the individuals, and update the positions of the stealing dung beetles. The formula is as follows:

[0124]

[0125] where x i (t) represents the position of the i-th dung beetle individual at the t-th generation; X * represents the local best position of the current dung beetle individual; X mean the average position of the dung beetle individuals; X b represents the current best dung beetle population

[0126] Step 6: Determine whether the termination condition is satisfied. If it is satisfied, stop the iteration and output the global best position and its fitness value; if not, re-execute Steps 1 to 5 until the termination condition is satisfied.

[0127] Step S344: Determine the corresponding parameter information based on the optimal dung beetle population position information, and use the corresponding parameter information as the optimized hyperparameters.

[0128] It can be understood that in this step, the optimal dung beetle population position information is used as the optimized hyperparameter, and the stability prediction model can more accurately approximate the global optimal solution, improving the accuracy and efficiency of the solution.

[0129] Step S35: Input the optimized hyperparameter into the fully connected layer for mapping to obtain the slope safety factor prediction model after training and optimization.

[0130] It can be understood that in this step, by optimizing the hyperparameter, the accuracy of the neural network prediction and the calculation speed are improved.

[0131] Step S4: Test the stability prediction model based on the test set to obtain the slope safety factor prediction model after testing;

[0132] It can be understood that in this step, by testing with the test set, the accuracy of the stability prediction is improved. In this step, step S4 includes step S41, step S42, step S43, step S44, and step S45.

[0133] Step S41: Substitute the optimized hyperparameter into the slope safety factor prediction model as the hidden state parameter of the reverse hidden layer to obtain the slope safety factor prediction model after substituting the optimized hyperparameter;

[0134] It can be understood that substituting the optimized hyperparameter into the slope safety factor prediction model improves the accuracy of the neural network prediction and the calculation speed.

[0135] Step S42: Input the training set into the slope safety factor prediction model after substituting the optimized hyperparameter for verification to obtain a five-fold cross-validation model;

[0136] It can be understood that in this step, the training set is first input into five slope safety factor prediction models after substituting the optimized hyperparameter to obtain a five-fold cross-validation model, and then the five-fold cross-validation model can be used for verification.

[0137] Step S43: Input the test set into the five-fold cross-validation model for prediction to obtain five slope safety factor prediction results;

[0138] It can be understood that in this step, the test set is respectively input into the five-fold cross-validation model for test verification, improving the accuracy of the verification, reducing the randomness of a single model, and ensuring the accuracy of the prediction model.

[0139] Step S44: Calculate the average value of all slope safety factor prediction results to obtain the final safety factor of the slope;

[0140] It can be understood that in this step, the root mean square error (RMSE) is used to evaluate the relative error between the predicted value and the true value, and the calculation formula is as follows:

[0141]

[0142] In the formula: y' t is the predicted value of the slope safety factor, y t is the true value of the slope safety factor, and n is the number of samples in the test set.

[0143] Step S45: Calculate the relative error between the final safety factor of the slope and the safety factor in the test set. When the relative error meets the accuracy requirement, the predicted model of the slope safety factor after testing is obtained.

[0144] It can be understood that in this step, a threshold is set to determine whether the error meets the accuracy requirement. If the error is less than or equal to the threshold, the accuracy requirement is met.

[0145] Step S5: Send the preset real-time monitoring data to the slope safety factor prediction model for prediction, obtain the safety factor corresponding to each real-time monitoring data, and use the safety factor corresponding to each real-time monitoring data as the slope stability value of the reverse wrapping geotextile.

[0146] It can be understood that in this step, by bringing the real-time monitoring data into the slope safety factor prediction model, the stability of the slope at a future time can be quickly predicted, providing a basis for judging whether the slope needs to be reinforced. In this step, step S5 includes step S51, step S52, and step S53.

[0147] Step S51: Generate a random field image from the real-time monitoring data, and generate a preset number of calculation samples based on the random field image. The number of the calculation samples is greater than the number of real-time monitoring data;

[0148] It can be understood that in this step, a random field image is generated from the real-time monitoring data, and then the random field images are randomly combined to generate N s calculation samples. The number of the calculation samples is greater than the number of data samples. Preferably, the N s = 50000;

[0149] Step S52: Input all the calculation samples into the slope safety factor prediction model in sequence for calculation, and predict the safety factor corresponding to each calculation sample;

[0150] It can be understood that in this step, by bringing the real-time monitoring data into the slope safety factor prediction model, the stability of the slope at a future time can be quickly predicted, providing a basis for judging whether the slope needs to be reinforced.

[0151] Step S53: Calculate the average value of the safety factors corresponding to all calculation samples to obtain the safety factor corresponding to the real-time monitoring data.

[0152] It can be understood that the calculation formula corresponding to the safety factor corresponding to the real-time monitoring data in this step is as follows:

[0153]

[0154] In the formula: F s,i is the predicted value of the safety factor corresponding to the real-time monitoring data, N s is the number of samples generated by the random combination of random fields.

[0155] Embodiment 2:

[0156] As Figure 2 shown, this embodiment provides a stability prediction system for an inverted-wrapped geotextile slope. Refer to Figure 2 The system includes an acquisition unit 701, a calculation unit 702, a training unit 703, a testing unit 704, and a prediction unit 705.

[0157] The acquisition unit 701 is configured to acquire the design data and historical monitoring data of the inverted-wrapped geotextile slope, and build a FreeMat mathematical optimization model based on the design data and historical monitoring data;

[0158] The calculation unit 702 is configured to input the design data and historical monitoring data of the inverted-wrapped geotextile slope into the FreeMat mathematical optimization model for stability calculation, and generate a data set based on the calculated safety factor data, design data, and historical monitoring data. The data set includes a training set and a testing set;

[0159] The training unit 703 is configured to train and optimize a preset neural network based on a preset improved dung beetle optimization algorithm and the training set to obtain a slope safety factor prediction model;

[0160] The testing unit 704 is configured to test the stability prediction model based on the testing set to obtain a slope safety factor prediction model after testing;

[0161] The prediction unit 705 is configured to send preset real-time monitoring data to the slope safety factor prediction model for prediction, obtain the safety factor corresponding to each real-time monitoring data, and use the safety factor corresponding to each real-time monitoring data as the stability value of the inverted-wrapped geotextile slope.

[0162] It should be noted that regarding the system in the above embodiment, the specific manners in which each module performs operations have been described in detail in the embodiment related to the method, and will not be elaborated here.

[0163] The above are only the preferred embodiments of the present invention and are not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

[0164] As described above, this is only the specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention can easily think of changes or replacements, which should all be covered within the protection scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.

Claims

1. A method for predicting the stability of an inverted geotextile slope, characterized in that: include: Obtain design data and historical monitoring data of the reverse geotextile slope, and build a FreeMat mathematical optimization model based on the design data and historical monitoring data; Input the design data and historical monitoring data of the reverse geotextile slope into the FreeMat mathematical optimization model to perform stability calculation, and generate a data set based on the calculated safety factor data, design data and historical monitoring data, wherein the data set includes a training set and a test set; Based on the preset improved dung beetle optimization algorithm and the training set, the preset neural network is trained and optimized to obtain a slope safety factor prediction model; Testing the stability prediction model based on the test set to obtain a slope safety factor prediction model after the test; The preset real-time monitoring data is sent to the slope safety factor prediction model for prediction, and the safety factor corresponding to each real-time monitoring data is obtained, and the safety factor corresponding to each real-time monitoring data is used as the anti-wrapped geotextile slope stability value.

2. The stability prediction method of the geotextile slope according to claim 1 is characterized in that ,The design data and historical monitoring data of the geotextile slope are input into the FreeMat mathematical optimization model for stability calculation, including: Determine the slope strip width and the number of control points based on the design data and historical monitoring data, and construct a horizontal coordinate matrix based on the slope strip width and the number of control points; Calculating the ordinate value of each control point based on the abscissa matrix and a preset sliding surface equation; Calculate the strip bottom inclination angles of a preset number of strip blocks based on the horizontal coordinate matrix and the vertical coordinate value of each control point to obtain a strip bottom inclination angle matrix; The safety factor value of the slope is calculated based on the design data and historical monitoring data, the abscissa matrix, the ordinate value of each control point and the strip bottom inclination matrix.

3. The stability prediction method of the geotextile slope according to claim 1 is characterized in that , based on the preset improved dung beetle optimization algorithm and the training set, the preset neural network is trained and optimized, including: Build a bidirectional recurrent neural network model based on the input layer, forward hidden layer, reverse hidden layer, fully connected layer and output layer; Use standard normal distribution to initialize the weights and parameters of the bidirectional recurrent neural network unit to obtain initialization parameters; Perform time modeling based on bidirectional recurrent neural network units and hidden layers, and output the hidden state at each moment; Taking the hidden state parameters of the reverse hidden layer in the bidirectional recurrent neural network model as hyperparameters, optimizing the hyperparameters based on a preset improved dung beetle optimization algorithm to obtain optimized hyperparameters; The optimized hyperparameters are input into the fully connected layer for mapping to obtain the trained and optimized slope safety factor prediction model.

4. The stability prediction method of the geotextile slope according to claim 3 is characterized in that , optimizing the hyper parameters based on the preset improved dung beetle optimization algorithm, including: Initialize all hyperparameters as input data to obtain the initialized dung beetle population; The average differential mutation perturbation, Levy-Cauchy mutation and greedy selection strategy were used to screen the dung beetle population and the screened dung beetle population was obtained. Calculating the fitness value of the screened dung beetle population based on a preset fitness function, and updating the position of the dung beetle population based on the fitness value until the number of iterations reaches a maximum number of iterations, and obtaining the optimal dung beetle population as position information; Based on the optimal dung beetle population location information, corresponding parameter information is determined, and the corresponding parameter information is used as the optimized hyperparameter.

5. The stability prediction method of the geotextile slope according to claim 4 is characterized in that ,Based on the test set, the stability prediction model is tested to obtain the slope safety factor prediction model after the test, including: Bringing the optimized hyperparameters into the slope safety factor prediction model as hidden state parameters of the reverse hidden layer therein, to obtain the slope safety factor prediction model after the optimized hyperparameters are brought into the model; The training set is input into the slope safety factor prediction model after the optimized hyperparameters are introduced for verification, and a five-fold cross-validation model is obtained; The test set is input into a five-fold cross validation model for prediction, and five slope safety factor prediction results are obtained; Calculate the average of all slope safety factor prediction results to get the final slope safety factor; The relative error between the final safety factor of the slope and the safety factor in the test set is calculated. When the relative error meets the accuracy requirement, the safety factor prediction model of the slope after the test is obtained.

6. A stability prediction system for geotextile slopes, characterized in that: include: An acquisition unit is used to acquire design data and historical monitoring data of the reverse geotextile slope, and to build a FreeMat mathematical optimization model based on the design data and historical monitoring data; A calculation unit, used for inputting the design data and historical monitoring data of the reverse geotextile slope into the FreeMat mathematical optimization model for stability calculation, and generating a data set based on the calculated safety factor data, design data and historical monitoring data, wherein the data set includes a training set and a test set; A training unit, used for training and optimizing a preset neural network based on a preset improved dung beetle optimization algorithm and the training set to obtain a slope safety factor prediction model; A testing unit, used for testing the stability prediction model based on the test set to obtain a slope safety factor prediction model after the test; The prediction unit is used to send the preset real-time monitoring data to the slope safety factor prediction model for prediction, obtain the safety factor corresponding to each real-time monitoring data, and use the safety factor corresponding to each real-time monitoring data as the anti-wrapped geotextile slope stability value.

7. The stability prediction system of the reverse geotextile slope according to claim 6, characterized in that: The computing unit comprises: A first calculation subunit is used to determine the width of the slope strips and the number of control points based on the design data and the historical monitoring data, and to construct a horizontal coordinate matrix based on the width of the slope strips and the number of control points; A second calculation subunit, used for calculating the ordinate value of each control point based on the abscissa matrix and a preset sliding surface equation; A third calculation subunit is used to calculate the strip bottom inclination angles of a preset number of strip blocks based on the horizontal coordinate matrix and the vertical coordinate value of each control point to obtain a strip bottom inclination angle matrix; The fourth calculation subunit is used to calculate the safety factor value of the slope based on the design data and historical monitoring data, the abscissa matrix, the ordinate value of each control point and the strip bottom inclination matrix.

8. The stability prediction system of the reverse geotextile slope according to claim 6, characterized in that: The training unit comprises: A first training subunit is used to build a bidirectional recurrent neural network model based on an input layer, a forward hidden layer, a reverse hidden layer, a fully connected layer, and an output layer; The second training subunit is used to initialize the weights and parameters of the bidirectional recurrent neural network unit using a standard normal distribution to obtain initialization parameters; The third training subunit is used for performing time modeling according to the bidirectional recurrent neural network unit and the hidden layer, and outputting the hidden state at each moment; A fourth training subunit is used to use the hidden state parameters of the reverse hidden layer in the bidirectional recurrent neural network model as hyperparameters, and optimize the hyperparameters based on a preset improved dung beetle optimization algorithm to obtain optimized hyperparameters; The fifth training subunit is used to input the optimized hyperparameters into the fully connected layer for mapping, so as to obtain the trained and optimized slope safety factor prediction model.

9. The stability prediction system of the reverse geotextile slope according to claim 8, characterized in that: The fourth training subunit comprises: The sixth training subunit is used to initialize all hyperparameters as input data to obtain an initialized dung beetle population; The seventh training subunit is used to screen the dung beetle population by using average differential mutation perturbation, Levy-Cauchy mutation and greedy selection strategy to obtain the screened dung beetle population; An eighth training subunit is used to calculate the fitness value of the screened dung beetle population based on a preset fitness function, and update the position of the dung beetle population based on the fitness value until the number of iterations reaches a maximum number of iterations, thereby obtaining the optimal dung beetle population as position information; The ninth training subunit is used to determine the corresponding parameter information based on the optimal dung beetle population location information, and use the corresponding parameter information as the optimized hyperparameter.

10. The stability prediction system of the reverse geotextile slope according to claim 9, characterized in that: The testing unit comprises: A first testing subunit is used to introduce the optimized hyperparameters into the slope safety factor prediction model as hidden state parameters of the reverse hidden layer therein, so as to obtain the slope safety factor prediction model after introducing the optimized hyperparameters; The second testing subunit is used to input the training set into the slope safety factor prediction model after the optimized hyperparameters are introduced for verification, so as to obtain a five-fold cross-validation model; A third testing subunit is used to input the test set into a five-fold cross validation model for prediction to obtain five slope safety factor prediction results; The fourth test subunit is used to calculate the average value of all slope safety factor prediction results to obtain the final safety factor of the slope; The fifth test subunit is used to calculate the relative error between the final safety factor of the slope and the safety factor in the test set. When the relative error meets the accuracy requirement, the slope safety factor prediction model after the test is obtained.

Images