Soft soil shield tunnel stability prediction method based on improved GA-BPNN

Abstract

The invention belongs to the technical field of soft soil shield tunnel stability prediction, and particularly discloses a soft soil shield tunnel stability prediction method based on improved GA-BPNN, and the method comprises the steps: determining initial stratum parameters, and generating a plurality of stratum parameter combinations through an orthogonal test method; inputting a plurality of groups of stratum parameter combinations into the finite element model, outputting to obtain tunnel settlement under different parameter combinations, and further constructing an improved GA-BPNN model; performing inversion analysis on the formation parameters based on the actual monitoring data and the improved GA-BPNN model to obtain updated formation parameters; and on the basis of the updated stratum parameters and the improved GA-BPNN model, the settlement amount of the soft soil shield tunnel is predicted, and the long-term stability of the soft soil shield tunnel is evaluated according to a prediction result. The problems that in the prior art, the precision and efficiency of parameter inversion and stability prediction of the soft soil shield tunnel cannot be guaranteed, and calculation resource consumption is large are solved.

Description

Technical Field

[0001] The present invention belongs to the technical field of soft soil shield tunnel stability prediction, and particularly relates to a soft soil shield tunnel stability prediction method based on improved GA - BPNN. Background Technique

[0002] With the acceleration of the urbanization process, the development and utilization of underground space have become an important part of urban infrastructure construction. As an important part of underground traffic engineering, shield tunnels often face complex geological conditions during the construction process. Especially in soft soil areas, the problem of tunnel settlement is particularly prominent, directly affecting construction safety and operation stability.

[0003] In the prior art, the settlement prediction of shield tunnels mostly relies on the formation parameters obtained from laboratory tests or in - situ tests. However, due to the limitations of environmental disturbance and the accuracy of testing instruments, the actual formation parameters obtained often have errors, resulting in insufficient accuracy of the prediction results. To solve the above problems, in recent years, artificial intelligence algorithms have been introduced into parameter inversion and settlement prediction, such as artificial neural network (BPNN), genetic algorithm (GA), etc. However, the limitations of single algorithms are relatively significant: BPNN is prone to falling into local optimum and has low training efficiency; although GA has good global search ability, it consumes a large amount of computing resources and has low accuracy. Summary of the Invention

[0004] The purpose of the present invention is to solve the problem that the prior art cannot guarantee the accuracy and efficiency of parameter inversion and stability prediction of soft soil shield tunnels, and consumes a large amount of computing resources, and proposes a soft soil shield tunnel stability prediction method based on improved GA - BPNN.

[0005] The technical solution of the present invention is as follows: A soft soil shield tunnel parameter inversion and stability prediction method based on improved GA - BPNN, comprising the following steps:

[0006] S1. Obtain the settlement data of the soft soil shield tunnel, then determine the initial formation parameters, and use numerical simulation software to establish a three - dimensional geological model of the soft soil shield tunnel construction;

[0007] S2. Based on the initial formation parameters, use the orthogonal test method to generate multiple groups of formation parameter combinations, and input the multiple groups of formation parameter combinations into the three - dimensional geological model of the soft soil shield tunnel construction, and output the tunnel settlement amounts under different parameter combinations;

[0008] S3. Construct an improved GA - BPNN model according to the tunnel settlement amounts under different parameter combinations;

[0009] S4. Based on the actual monitoring data and the improved GA - BPNN model, conduct inversion analysis on the formation parameters to be processed to obtain updated formation parameters;

[0010] S5. Based on the updated formation parameters and the improved GA-BPNN model, predict the settlement of the soft soil shield tunnel, and evaluate the long-term stability of the soft soil shield tunnel according to the prediction results.

[0011] The beneficial effects of the present invention are as follows:

[0012] 1. By generating multiple groups of formation parameter combinations through the orthogonal experimental design method, the calculation cost can be reduced and the variation range of formation parameters can be fully covered;

[0013] 2. By combining on-site monitoring data and finite element numerical simulation, the present invention uses the improved GA-BPNN model to perform inverse analysis on the mechanical parameters of the soft soil shield tunnel and predict the tunnel stability, which can effectively improve the accuracy of parameter inversion and provide theoretical and practical basis for the safety of the soft soil shield tunnel.

[0014] Preferably, the settlement data of the soft soil shield tunnel includes settlement curves, formation deformation amounts, and relevant construction parameters;

[0015] The initial formation parameters include unit weight, Poisson's ratio, internal friction angle, cohesion, and modulus.

[0016] Preferably, the step S1 specifically includes the following sub-steps:

[0017] Set monitoring points to collect the settlement data of the soft soil shield tunnel in real time, and the monitoring points are evenly distributed on the left and right lines of the soft soil shield tunnel;

[0018] According to the geological exploration data of the project area, establish a formation distribution model, and extract the initial physical and mechanical parameters of each formation, that is, the initial formation parameters, based on the exploration data;

[0019] Taking the soft soil shield tunnel as the core, use finite element numerical simulation software to establish a three-dimensional geological model of the soft soil shield tunnel construction.

[0020] Preferably, the step S3 specifically includes the following sub-steps:

[0021] S31. Construct a BP neural network model;

[0022] S32. According to the tunnel settlement amounts under different parameter combinations, use the genetic algorithm to optimize the initial weights and bias values of the BP neural network to obtain an optimized BP neural network;

[0023] S33. Train the optimized BP neural network and use the sensitivity factor to adjust the fitness function of the optimized BP neural network to obtain an improved GA-BPNN model.

[0024] Preferably, the BP neural network model in step S3 includes an input layer, a hidden layer, and an output layer connected in sequence;

[0025] The input data of the input layer is formation parameters;

[0026] The hidden layer is used to receive the output data of the input layer, extract non-linear features from the output data of the input layer, and output non-linear features;

[0027] The output layer is used to receive and process the non-linear features, and output the settlement prediction value of the soft soil shield tunnel;

[0028] The calculation formula for the number of neurons in the hidden layer is:

[0029] m = n0log2n′, n0 ∈ [1, 10]

[0030] Among them, m represents the number of neurons in the hidden layer, n0 represents the correction coefficient, log represents the logarithmic function with base 2, and n′ represents the number of neurons in the input layer.

[0031] Preferably, step S32 specifically includes the following sub-steps:

[0032] S321. Take the tunnel settlement amounts under different parameter combinations as the labels of the BP neural network model, encode the weights and biases of the BP neural network model to obtain an initial population;

[0033] S322. Determine the fitness function;

[0034] S323. Iteratively update the population through selection, crossover, and mutation operations;

[0035] S324. Calculate the fitness according to the fitness function;

[0036] S325. Judge whether the maximum number of iterations is reached or the fitness reaches a preset threshold. If so, obtain the optimal population, that is, the optimal weights and biases of the BP neural network model, and obtain the optimized BP neural network; if not, return to step S322.

[0037] Preferably, the calculation formula for the sensitivity factor in step S33 is:

[0038]

[0039] Among them, S represents the sensitivity factor, Y i represents the output value of the i-th iterative calculation, Y0 represents the output value calculated based on the initial parameters, P represents the percentage change of the parameter value used in the i-th iteration relative to the initial parameters, q represents the number of iterations, Y i+1represents the output value of the (i + 1)-th iterative calculation, P i+1 represents the percentage change of the parameter value used in the (i + 1)-th iteration relative to the initial parameter.

[0040] Preferably, the calculation formula of the fitness function for adjusting and optimizing the BP neural network with the sensitivity factor in step S33 is:

[0041]

[0042] where F(x) represents the fitness function of the BP neural network adjusted and optimized by the sensitivity factor, A hl represents the actual settlement monitoring value of the tunnel, E hl represents the predicted value of the actual settlement of the tunnel, S hl represents the sensitivity factor, H represents the number of parameters for inverse analysis, that is, the number of formation parameters, and L represents the actual monitoring data.

[0043] Preferably, step S4 specifically includes the following steps:

[0044] Input the real-time monitored settlement data of the soft soil shield tunnel into the improved GA - BPNN model, and output the predicted settlement value;

[0045] According to the predicted settlement value, adjust the formation parameter combination of the formation parameters to be processed, so as to minimize the error between the settlement value predicted by the improved GA - BPNN model and the actual monitoring value, and then inversely obtain the updated formation parameters.

[0046] The beneficial effects of the above preferred solution are:

[0047] By optimizing the initial weights and biases of the BPNN through the genetic algorithm, and combining the introduction of the sensitivity factor and the finite element numerical simulation method, the accuracy and efficiency of inversion and prediction are significantly improved, providing theoretical and practical support for the construction safety of soft soil shield tunnels. Brief Description of the Drawings

[0048] Figure 1 Shown is a flowchart of a method for inverse analysis of parameters and stability prediction of a soft soil shield tunnel based on an improved GA - BPNN provided by an embodiment of the present invention.

[0049] Figure 2 Shown is a schematic diagram of a three - dimensional geological model for the construction of a soft soil shield tunnel provided by an embodiment of the present invention.

[0050] Figure 3 Shown is a sectional view of the three - dimensional geological model for the construction of a soft soil shield tunnel provided by an embodiment of the present invention.

[0051] Figure 4The following is a flowchart of the parameter inversion analysis based on the improved GA - BPNN model provided by the embodiments of the present invention. Detailed implementation manners

[0052] Now, the exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings. It should be understood that the embodiments shown and described in the drawings are merely exemplary, intended to illustrate the principles and spirit of the present invention, rather than limiting the scope of the present invention.

[0053] As Figure 1 shown, a method for parameter inversion and stability prediction of soft - soil shield tunnels based on an improved genetic algorithm and back - propagation neural network (GA - BPNN) includes the following steps:

[0054] S1. Obtain the settlement data of the soft - soil shield tunnel, then determine the initial formation parameters, and use numerical simulation software to establish a three - dimensional geological model of the soft - soil shield tunnel construction;

[0055] S2. Based on the initial formation parameters, use the orthogonal test method to generate multiple groups of formation parameter combinations, and input the multiple groups of formation parameter combinations into the three - dimensional geological model of the soft - soil shield tunnel construction, and output the tunnel settlement amounts under different parameter combinations;

[0056] To reduce the calculation cost and fully cover the variation range of soil parameters, the orthogonal experimental design method is used to generate soil samples with different parameter combinations. The parameters that mainly affect the shield settlement are used as variables, and their reasonable variation ranges are set, and multiple groups of parameter combinations are generated by orthogonal design;

[0057] Substitute the formation parameter combinations generated by orthogonal design into the numerical model to simulate the settlement amounts under different parameter conditions during the shield tunnel construction. The simulation results form the training data set and test data set of parameter - settlement;

[0058] S3. Construct an improved GA - BPNN model according to the tunnel settlement amounts under different parameter combinations;

[0059] S4. Based on the actual monitoring data and the improved GA - BPNN model, conduct an inversion analysis on the formation parameters to be processed to obtain updated formation parameters;

[0060] S5. Based on the updated formation parameters and the improved GA - BPNN model, predict the settlement amount of the soft - soil shield tunnel, and evaluate the long - term stability of the soft - soil shield tunnel according to the prediction results;

[0061] Using the updated stratum parameters and combining with the improved GA - BPNN model to predict the settlement amount during the whole life cycle of the soft - soil shield tunnel. The prediction results are compared and verified with the on - site monitoring data to ensure the prediction accuracy. According to the predicted settlement amount, combined with the design safety standards and specifications of the tunnel structure, the long - term stability of the shield tunnel is evaluated. If the predicted settlement amount exceeds the safety threshold, a warning signal is sent, and it is recommended to adjust the construction parameters or take reinforcement measures.

[0062] In this embodiment, step S1 specifically includes the following sub - steps:

[0063] During the construction process of the soft - soil shield tunnel, monitoring points are set to collect the settlement data of the soft - soil shield tunnel in real - time. The monitoring points are evenly distributed on the left and right lines of the soft - soil shield tunnel. The settlement data of the soft - soil shield tunnel includes settlement curves, stratum deformation amounts, and related construction parameters.

[0064] According to the geological exploration data of the project area, a stratum distribution model is established. Taking a certain soft - soil shield tunnel project in a certain place as an example, the main strata in the research area include silt layers, silty clay layers, fine - medium sand layers, etc. Then, based on the exploration data, the initial physical and mechanical parameters of each stratum, that is, the initial stratum parameters, are extracted. The initial physical and mechanical parameters include unit weight, Poisson's ratio, internal friction angle, cohesion, and modulus, etc., as the initial input parameters for finite - element numerical simulation.

[0065] Use finite - element numerical simulation software to establish a three - dimensional geological model of the soft - soil shield tunnel construction, that is, a finite - element model, as shown in Figure 2 and Figure 3 The size of the three - dimensional geological model is set according to the actual engineering situation. With the shield tunnel as the core, the surrounding stratum soil covers the main geological layers. The model boundary needs to be far enough to avoid the influence of boundary effects on the calculation results. The finite - element numerical simulation software can be Plaxis 3D in this embodiment.

[0066] In this embodiment, step S3 specifically includes the following sub - steps:

[0067] S31. Construct a BP neural network model;

[0068] S32. According to the tunnel settlement amounts under different parameter combinations, use the genetic algorithm to optimize the initial weights and bias values of the BP neural network to obtain an optimized BP neural network.

[0069] S33. Train the optimized BP neural network and use the sensitivity factor to adjust the fitness function of the optimized BP neural network to obtain an improved GA - BPNN model. The sensitivity factor is used to measure the influence degree of each stratum parameter on the settlement prediction result. Based on the sensitivity factor to adjust the fitness function of the optimized BP neural network can further improve the prediction accuracy.

[0070] In this embodiment, the BP neural network model in step S3 includes an input layer, a hidden layer, and an output layer connected in sequence;

[0071] The input data of the input layer includes the tunnel settlement amount and formation parameters under different parameter combinations of the monitoring points;

[0072] The hidden layer is used to receive the output data of the input layer, perform non-linear feature extraction on the output data of the input layer, and output non-linear features;

[0073] The output layer is used to receive and process the non-linear features and output the settlement prediction value of the soft soil shield tunnel;

[0074] The calculation formula for the number of neurons in the hidden layer is:

[0075] m = n0log2n′, n0 ∈ [1, 10]

[0076] where m represents the number of neurons in the hidden layer, n0 represents the correction coefficient, log represents the logarithmic function with base 2, and n′ represents the number of neurons in the input layer.

[0077] In this embodiment, step S32 specifically includes the following sub-steps:

[0078] S321. Take the tunnel settlement amounts under different parameter combinations as the labels of the BP neural network model, encode the weights and biases of the BP neural network model to obtain an initial population;

[0079] S322. Determine the fitness function, and the fitness function takes the error reverse degree (such as the mean square error MSE) between the predicted value and the target value as the optimization objective;

[0080] S323. Iteratively update the population through selection, crossover, and mutation operations, and evolve generation by generation to approach the global optimal solution;

[0081] S324. Calculate the fitness according to the fitness function;

[0082] S325. Judge whether the maximum number of iterations is reached or the fitness reaches the preset threshold. If so, obtain the optimal population, that is, the optimal weights and biases of the BP neural network model, and obtain the optimized BP neural network; if not, return to step S322.

[0083] In this embodiment, the calculation formula for the sensitivity factor in step S33 is:

[0084]

[0085] where S represents the sensitivity factor, Yi represents the output value of the i-th iterative calculation, Y0 represents the output value calculated based on the initial parameters, P represents the percentage change of the parameter value used in the i-th iteration relative to the initial parameters, q represents the number of iterations, and Y i+1 represents the output value of the (i + 1)-th iterative calculation, and P i+1 represents the percentage change of the parameter value used in the (i + 1)-th iteration relative to the initial parameters.

[0086] In this embodiment, the calculation formula of the fitness function of the BP neural network optimized by adjusting the sensitivity factor in step S33 is:

[0087]

[0088] where F(x) represents the fitness function of the BP neural network optimized by adjusting the sensitivity factor, and A hl represents the actual settlement monitoring value of the tunnel, and E hl represents the predicted value of the actual settlement of the tunnel, and S hl represents the sensitivity factor, H represents the number of parameters for inverse analysis, that is, the number of formation parameters, and L represents the actual monitoring data.

[0089] In this embodiment, as Figure 4 shown, the step S4 specifically includes the following steps:

[0090] Input the real-time monitored settlement data of the soft soil shield tunnel into the improved GA - BPNN model, and output the predicted settlement value;

[0091] According to the predicted settlement value, adjust the formation parameter combination of the formation parameters to be processed to minimize the error between the settlement value predicted by the improved GA - BPNN model and the actual monitored value, and then inversely obtain the updated formation parameters.

[0092] The method for parameter inversion and stability prediction of soft soil shield tunnels based on the improved GA - BPNN proposed by the present invention optimizes the initial weights and biases of the BPNN through GA, and combines the introduction of sensitivity factors and the finite element numerical simulation method, significantly improving the accuracy and efficiency of inversion and prediction, and providing theoretical and practical support for the construction safety of soft soil shield tunnels.

[0093] Those of ordinary skill in the art will realize that the embodiments described here are for helping readers understand the principles of the present invention, and it should be understood that the protection scope of the present invention is not limited to such specific statements and embodiments. Those of ordinary skill in the art can make various specific deformations and combinations that do not depart from the essence of the present invention according to these technical revelations disclosed by the present invention, and these deformations and combinations are still within the protection scope of the present invention.

Claims

1. A method for inverse analysis of soft soil shield tunnel parameters and stability prediction based on improved GA-BPNN, characterized in that, The method includes the following steps: S1. Obtain the settlement data of the soft soil shield tunnel, and then determine the initial formation parameters, and use numerical simulation software to establish a three-dimensional geological model for the construction of the soft soil shield tunnel; S2. Based on the initial formation parameters, use the orthogonal test method to generate multiple groups of formation parameter combinations, and input the multiple groups of formation parameter combinations into the three-dimensional geological model for the construction of the soft soil shield tunnel, and output the tunnel settlement amounts under different parameter combinations; S3. Construct an improved GA-BPNN model according to the tunnel settlement amounts under different parameter combinations; S4. Based on the actual monitoring data and the improved GA-BPNN model, perform inversion analysis on the formation parameters to be processed to obtain updated formation parameters; S5. Based on the updated formation parameters and the improved GA-BPNN model, predict the settlement amount of the soft soil shield tunnel, and evaluate the long-term stability of the soft soil shield tunnel according to the prediction results.

2. The method for inverse parameter analysis and stability prediction of soft soil shield tunnel based on improved GA-BPNN according to claim 1, characterized in that The settlement data of the soft soil shield tunnel includes settlement curves, formation deformation amounts, and relevant construction parameters; The initial formation parameters include unit weight, Poisson's ratio, internal friction angle, cohesion, and modulus.

3. The method for inverse analysis of soft soil shield tunnel parameters and stability prediction based on improved GA-BPNN according to claim 2, characterized in that The specific steps of step S1 include the following sub-steps: Set monitoring points to collect the settlement data of the soft soil shield tunnel in real time, and the monitoring points are evenly distributed on the left and right lines of the soft soil shield tunnel; According to the geological exploration data of the project area, establish a formation distribution model, and based on the exploration data, extract the initial physical and mechanical parameters of each formation, that is, the initial formation parameters; Taking the soft soil shield tunnel as the core, use finite element numerical simulation software to establish a three-dimensional geological model for the construction of the soft soil shield tunnel.

4. The method for inverse analysis of soft soil shield tunnel parameters and stability prediction based on improved GA-BPNN according to claim 1, characterized in that, The specific steps of step S3 include the following sub-steps: S31. Construct a BP neural network model; S32. According to the tunnel settlement amounts under different parameter combinations, use the genetic algorithm to optimize the initial weights and bias values of the BP neural network to obtain an optimized BP neural network; S33. Train the optimized BP neural network, and use the sensitivity factor to adjust the fitness function of the optimized BP neural network to obtain an improved GA-BPNN model.

5. The method for inverse analysis of soft soil shield tunnel parameters and stability prediction based on improved GA-BPNN according to claim 4, characterized in that, The BP neural network model in step S3 includes an input layer, a hidden layer, and an output layer connected in sequence; The input data of the input layer is formation parameters; The hidden layer is used to receive the output data of the input layer, extract non-linear features from the output data of the input layer, and output non-linear features; The output layer is used to receive and process the non-linear features, and output the settlement prediction value of the soft soil shield tunnel; The calculation formula for the number of neurons in the hidden layer is: m = n0log2n′, n0 ∈ [1, 10] where m represents the number of neurons in the hidden layer, n0 represents the correction coefficient, log represents the logarithmic function with base 2, and n′ represents the number of neurons in the input layer.

6. The method for inverse analysis of soft soil shield tunnel parameters and stability prediction based on improved GA-BPNN according to claim 4, characterized in that The specific steps of step S32 include the following sub-steps: S321. Use the tunnel settlement amounts under different parameter combinations as the labels of the BP neural network model, encode the weights and biases of the BP neural network model to obtain an initial population; S322. Determine the fitness function; S323. Iteratively update the population through selection, crossover, and mutation operations; S324. Calculate the fitness according to the fitness function; S325. Determine whether the maximum number of iterations is reached or the fitness reaches the preset threshold. If so, obtain the optimal population, that is, the optimal weights and biases of the BP neural network model, and obtain the optimized BP neural network; if not, return to step S322.

7. The method for inverse analysis of soft soil shield tunnel parameters and stability prediction based on improved GA-BPNN according to claim 4, characterized in that The calculation formula for the sensitivity factor described in step S33 is: Among them, S represents the sensitivity factor, Y i represents the output value of the i-th iterative calculation, Y0 represents the output value calculated based on the initial parameters, P represents the percentage change of the parameter value used in the i-th iteration relative to the initial parameters, q represents the number of iterations, Y i+1 represents the output value of the (i + 1)-th iterative calculation, P i+1 represents the percentage change of the parameter value used in the (i + 1)-th iteration relative to the initial parameters.

8. The parameter inversion and stability prediction method of soft soil shield tunnel based on improved GA-BPNN according to claim 4, characterized in that, The calculation formula for the fitness function that the sensitivity factor described in step S33 adjusts and optimizes the BP neural network is: Among them, F(x) represents the fitness function of the BP neural network optimized by adjusting the sensitivity factor, A hl represents the actual settlement monitoring value of the tunnel, E hl represents the predicted value of the actual settlement of the tunnel, S hl represents the sensitivity factor, H represents the number of parameters in the back analysis, that is, the number of formation parameters, and L represents the actual monitoring data.

9. The method for inverse analysis of soft soil shield tunnel parameters and stability prediction based on improved GA-BPNN according to claim 1, characterized in that Step S4 specifically includes the following steps: Input the real-time monitored settlement data of the soft soil shield tunnel into the improved GA-BPNN model, and output the predicted settlement value; According to the predicted settlement value, adjust the formation parameter combination of the formation parameters to be processed to minimize the error between the settlement value predicted by the improved GA-BPNN model and the actual monitored value, and then inversely obtain the updated formation parameters.

Images