Method for acquiring mechanical parameters of tunnel surrounding rock

Abstract

The invention relates to a tunnel surrounding rock mechanical parameter acquisition method, which comprises the following steps of S1, selecting to-be-inverted rock mechanical parameters of a tunnel surrounding rock mass, and constructing a calculation scheme of a parameter sample by adopting a Latin hypercube sampling method; s2, taking the calculation scheme and the displacement amount of the corresponding hole periphery measuring point as an input parameter sample set and an output parameter sample set; s3, establishing a least square support vector machine (LSSVM) numerical agent model of a nonlinear mapping relation between the to-be-inverted surrounding rock mechanical parameter and the displacement of the hole periphery measuring point; s4, adopting an improved firefly algorithm to obtain an optimal IFA-LSSVM numerical agent model; and S5, substituting field actual measurement displacement data into the numerical agent model to obtain final rock and soil mass mechanical parameters. Compared with the prior art, the method has the advantages of improving the precision and accuracy of the LSSVM numerical agent model and the like.

Description

Technical Field

[0001] The invention relates to the fields of tunnel engineering and geotechnical engineering, and in particular to a method for obtaining mechanical parameters of tunnel surrounding rocks. Background Art

[0002] In engineering practice, the mechanical parameters of rock and soil are often obtained through on-site in-situ surveys and indoor tests. However, due to the characteristics of uneven spatial distribution and susceptibility to disturbance of rock and soil materials, general indoor tests and on-site in-situ surveys are limited by factors such as manpower, time, and cost, and cannot fully and effectively cover the construction area. Usually, there are only limited survey results in certain parts of the construction area. At the same time, due to the difference between the test environment and the actual stress environment, the rock mechanical parameters obtained by indoor tests are also difficult to accurately reflect the actual surrounding rock properties.

[0003] At present, the numerical calculation of stability analysis of tunnel and underground engineering and geotechnical engineering requires accurate rock mass mechanical parameters. However, due to the randomness, heterogeneity, nonlinearity, discontinuity of rock materials and the influence of various engineering and construction factors, it is almost impossible to solve them using analytical methods. It is difficult to make reasonable estimates of the input parameters of the numerical method calculation model. The rock mass mechanical parameters determined by indoor tests or field tests have large deviations from the actual parameters. Moreover, due to the influence of joints and cracks, the experimental results are not representative. When such parameters are used as calculation input parameters for numerical analysis, the results obtained often have large errors from the actual situation, which makes it difficult to use in engineering practice, and affects the further promotion and application of numerical methods in rock mass engineering to varying degrees. The displacement back analysis of rock mass mechanical parameters combines computer technology, numerical analysis methods, optimization design and field measurements to obtain more accurate calculation parameters of numerical simulation models.

[0004] At present, the displacement back analysis method based on construction monitoring data is commonly used to infer the surrounding rock parameters. For example, Chinese patent CN111414658A discloses a rock mass mechanical parameter back analysis method, which includes the following steps: S1, determining the rock mechanical parameters to be inverted, and constructing a calculation scheme for training samples based on the uniform test design method; S2, performing numerical calculations on each constructed scheme to obtain the valley deformation value corresponding to each scheme, and forming the calculation scheme and the corresponding valley deformation calculation value into the input and output values ​​of the SDCS-LSSVM algorithm; S3, based on the input and output samples obtained in step S2, and by learning the input and output sample data, establishing a nonlinear mapping relationship between rock mechanical parameters and valley deformation values; S4, solving the optimal solution of the objective function and determining the optimal combination of mechanical parameters.

[0005] The back analysis process of tunnel surrounding rock mechanical parameters and displacement requires a large number of forward analysis numerical calculations. Problems such as large amount of calculation, low efficiency, and poor solution stability will affect the accuracy of the final result. Using machine learning methods to establish a numerical proxy model of the nonlinear mapping relationship between surrounding rock mechanical parameters and displacement can significantly reduce the time cost of numerical calculations. However, there is currently little research on the back analysis of tunnel surrounding rock mechanical parameters based on machine learning proxy models.

[0006] The above method has the following defects: the above method does not establish a numerical proxy model, cannot significantly reduce the time cost of numerical calculation, and does not optimize the hyperparameters of the numerical proxy model, and cannot obtain the optimal numerical proxy model. Summary of the invention

[0007] The purpose of the present invention is to provide a method for obtaining mechanical parameters of tunnel surrounding rock in order to improve the precision and accuracy of the LSSVM numerical proxy model.

[0008] The purpose of the present invention can be achieved by the following technical solutions:

[0009] A method for obtaining mechanical parameters of tunnel surrounding rock, the method comprising the following steps:

[0010] S1: Select the rock mechanical parameters of the tunnel surrounding rock mass to be inverted and the calculation scheme of the construction parameter sample;

[0011] S2: Perform numerical calculation on each constructed calculation scheme to obtain the displacement of the hole surrounding measurement points corresponding to each calculation scheme, and use the calculation scheme and the displacement of the corresponding hole surrounding measurement points as input parameter and output parameter sample sets;

[0012] S3: Based on the input parameter and output parameter sample set obtained in step S2, a least squares support vector machine (LSSVM) numerical proxy model of the nonlinear mapping relationship between the surrounding rock mechanical parameters to be inverted and the displacement of the measuring points around the tunnel is established;

[0013] S4: The improved firefly algorithm IFA is used to obtain the optimal combination of the hyperparameter regularization parameter γ and the Gaussian kernel parameter σ of the LSSVM numerical proxy model, and then the optimal IFA-LSSVM numerical proxy model is obtained;

[0014] S5: Substitute the measured displacement data in the field into the optimal IFA-LSSVM numerical proxy model obtained in step S4 to obtain the final mechanical parameters of the rock and soil mass.

[0015] Furthermore, the specific steps of S1 are:

[0016] S1-1: Select the mechanical parameters to be inverted by principal component analysis or factor sensitivity analysis;

[0017] S1-2: Determine the value range of the rock mechanics parameters to be inverted for the tunnel surrounding rock mass based on the on-site investigation of the tunnel site;

[0018] S1-3: Based on the value range, the Latin hypercube sampling method is used to construct a calculation scheme for parameter samples.

[0019] Furthermore, the specific steps of S4 are:

[0020] S4-1: Construct an improved firefly algorithm;

[0021] S4-2: Using the improved firefly algorithm of S4-1, the hyperparameter regularization parameter γ and Gaussian kernel parameter σ of the LSSVM numerical proxy model are optimized and calculated to obtain the optimal combination of the hyperparameter regularization parameter γ and the Gaussian kernel parameter σ. The optimal combination is substituted into the LSSVM numerical proxy model to obtain the optimal IFA-LSSVM numerical proxy model.

[0022] Furthermore, the steps of S4-2 are:

[0023] Initialize the position of the firefly;

[0024] Calculate the fitness function of the firefly according to the current position of the firefly;

[0025] Update the position of the firefly. The updated position of the firefly is:

[0026] x j+1 =β(t)x j +(1-β(t))x i +α(t)ε j

[0027] Among them, α(t) is the adaptive moving step size; is the attractiveness value; β 0 is the attraction value at the light source (r = 0); γ t is the adaptive light absorption coefficient; ε j is a random number vector that follows a uniform or normal distribution; r ij represents the Cartesian distance between any two firefly individuals i and j;

[0028] Iterate the calculation until the number of iterations reaches the threshold, output the optimal combination of the hyperparameter regularization parameter γ and the Gaussian kernel parameter σ, substitute the optimal combination into the LSSVM numerical proxy model, and obtain the optimal IFA-LSSVM numerical proxy model.

[0029] Furthermore, the adaptive moving step size is:

[0030]

[0031] Among them, α 0 is the initial moving step length, Itr max is the maximum number of iterations, c is an integer that determines the randomness decay rate, and t is the current number of iterations.

[0032] Furthermore, the adaptive light absorption coefficient is:

[0033]

[0034] Among them, γ 0 is the initial light intensity absorption coefficient.

[0035] Furthermore, the calculation scheme for constructing parameter samples adopts the Latin hypercube sampling method.

[0036] Furthermore, rock mechanical parameters include elastic modulus, Poisson's ratio, internal friction angle and cohesion.

[0037] Furthermore, rock mechanical parameters are selected based on principal component analysis or factor sensitivity analysis methods.

[0038] Furthermore, the finite element method is used for numerical calculations.

[0039] Compared with the prior art, the present invention has the following beneficial effects:

[0040] (1) In the standard firefly algorithm (FA), many parameter values ​​in the position update equation, such as the moving step length α, are predetermined and fixed. The present invention obtains the optimal combination of the hyperparameter regularization parameter γ and the Gaussian kernel parameter σ of the LSSVM numerical proxy model, and the light intensity absorption coefficient γ of the firefly algorithm is t The moving step α is modified so that it can adaptively change the value with the number of iterations, so that the algorithm can have a better convergence speed in the early stage, enhance the search for local extreme values, and avoid jumping out of the local search in the later stage of the search. Finally, the algorithm can quickly obtain the optimal solution of the parameters of the LSSVM numerical proxy model with better accuracy, improve the precision and accuracy, significantly reduce the time cost of numerical calculation, and have higher inversion efficiency. .

[0041] (2) The Latin hypercube sampling method adopted in the present invention performs better in terms of uniform distribution and reducing the number of iterations, and is suitable for large-scale data processing. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 It is a flow chart of the present invention. DETAILED DESCRIPTION

[0043] The present invention is described in detail below in conjunction with the accompanying drawings and specific embodiments. This embodiment is implemented based on the technical solution of the present invention, and provides a detailed implementation method and specific operation process, but the protection scope of the present invention is not limited to the following embodiments.

[0044] The present invention proposes a method for obtaining mechanical parameters of tunnel surrounding rock, the flow chart of which is as follows: Figure 1 The method comprises the following steps: S1: selecting rock mechanical parameters of the tunnel surrounding rock mass to be inverted and constructing a calculation scheme for parameter samples;

[0045] S2: Perform numerical calculation on each constructed calculation scheme to obtain the displacement of the hole surrounding measurement points corresponding to each calculation scheme, and use the calculation scheme and the displacement of the corresponding hole surrounding measurement points as input parameter and output parameter sample sets;

[0046] S3: Based on the input parameter and output parameter sample set obtained in step S2, a least squares support vector machine (LSSVM) numerical proxy model of the nonlinear mapping relationship between the surrounding rock mechanical parameters to be inverted and the displacement of the measuring points around the tunnel is established;

[0047] S4: The improved firefly algorithm IFA is used to obtain the optimal combination of the hyperparameter regularization parameter γ and the Gaussian kernel parameter σ of the LSSVM numerical proxy model, and then the optimal IFA-LSSVM numerical proxy model is obtained;

[0048] S5: Substitute the measured displacement data in the field into the optimal IFA-LSSVM numerical proxy model obtained in step S4 to obtain the final mechanical parameters of the rock and soil mass.

[0049] Step S1 is specifically as follows:

[0050] S1-1: Select the mechanical parameters to be inverted by principal component analysis or factor sensitivity analysis;

[0051] S1-2: Determine the value range of the rock mechanics parameters to be inverted of the tunnel surrounding rock mass based on the actual conditions such as the on-site investigation of the tunnel site;

[0052] S1-3: Computational scheme for constructing parameter samples using the Latin hypercube sampling method.

[0053] Step S2 is specifically as follows:

[0054] S2-1: Use finite element software ABAQUS to perform numerical calculations on each calculation scheme constructed in S1-3, and obtain the displacement of the measuring points around the tunnel corresponding to each calculation scheme;

[0055] S2-2: Each calculation scheme is combined with the corresponding surrounding rock displacement calculation value obtained in S2-1 to form the overall data set of input parameters and output parameters of the IFA-LSSVM algorithm.

[0056] In step 3, based on the input parameter and output parameter sample set obtained in step S2, a least squares support vector machine (LSSVM) numerical proxy model of the nonlinear mapping relationship between the surrounding rock mechanical parameters to be inverted and the displacement of the measuring points around the tunnel is established.

[0057] Step S4 is specifically as follows:

[0058] S4-1: Improved Firefly Algorithm (IFA).

[0059] More preferably, the improved firefly algorithm (IFA) is specifically:

[0060] The mathematical expression for updating the firefly position in the improved firefly algorithm (IFA) is:

[0061] x j+1 =β(t)x j +(1-β(t))x i +α(t)ε j

[0062] Where α(t) is the adaptive moving step size; is the attractiveness value; β 0 is the attraction value at the light source (r = 0); γ t is the adaptive light absorption coefficient; ε j is a random number vector that follows a uniform or normal distribution; r ij Represents the Cartesian distance between any two firefly individuals i and j.

[0063] More preferably, the improvements α and γ are specifically:

[0064]

[0065] In the formula, α 0 and γ 0 is the initial moving step length and the initial light intensity absorption coefficient; c is an integer that determines the random decay rate, generally c = 5; Itr max Itr is the maximum number of iterations; i is the current iteration number.

[0066] S4-2: Using the improved firefly algorithm (IFA) in S4-1, the hyperparameter regularization parameter γ and Gaussian kernel parameter σ of the LSSVM numerical proxy model established in S3 are optimized and calculated to obtain the optimal combination of the hyperparameter regularization parameter γ and the Gaussian kernel parameter σ, and then the optimal IFA-LSSVM numerical proxy model is obtained.

[0067] Preferably, in step S5, the field measured displacement data is substituted into the optimal IFA-LSSVM numerical proxy model obtained in step S4-2 to obtain the final rock and soil mechanical parameters.

[0068] Compared with the prior art, the present invention has the following beneficial effects:

[0069] (1) In the standard firefly algorithm (FA), many parameter values ​​in the position update equation, such as the moving step length α, are predetermined and fixed. t The moving step size α is modified so that it can adaptively change its value with the number of iterations, so that the algorithm can have a better convergence speed in the early stage, enhance the search for local extreme values, and avoid jumping out of the local search in the later stage of the search, finally enabling the algorithm to quickly obtain the optimal solution with better accuracy.

[0070] (2) The Latin hypercube sampling method adopted in the present invention performs better in terms of uniform distribution and reducing the number of iterations, and is suitable for large-scale data processing.

[0071] (3) High efficiency: The present invention establishes the optimal IFA-LSSVM numerical proxy model, which can significantly reduce the time cost of numerical calculation and has higher inversion efficiency.

[0072] The cross section of a certain tunnel is a three-center circle, with a height of 10.8m, a span of 12m, a burial depth of 40.5m, and a Class IV surrounding rock. The surrounding rock parameters of the tunnel have a large range of variation. In order to ensure safe and reliable construction, the present invention is used to invert the surrounding rock parameters based on the monitoring data. The range of surrounding rock parameter variation is shown in Table 1.

[0073] Table 1. Range of values ​​of parameters to be inverted

[0074]

[0075] According to the range of values ​​shown in Table 1, 1000 sets of surrounding rock parameter data were generated by Latin hypercube random sampling, and the ABAQUS model of the tunnel was established to calculate the displacement values ​​of each measuring point. The IFA-LSSVM algorithm of the improved firefly algorithm (IFA)-least squares support vector machine (LSSVM) of the established data set was studied to obtain the optimal IFA-LSSVM numerical proxy model. Then, the measured displacement data on site was substituted into the optimal IFA-LSSVM numerical proxy model obtained in step S4 to obtain the final rock and soil mechanical parameters as shown in Table 2. Finally, the inversion results of the surrounding rock mechanical parameters were substituted into the finite element numerical simulation to obtain the calculated values ​​of the displacement and the measured values, as shown in Table 3.

[0076] Table 2 Inversion results of surrounding rock mechanical parameters

[0077]

[0078] Table 3 Comparison of displacement calculation values ​​corresponding to inversion results and actual monitoring values

[0079]

[0080] The preferred specific embodiments of the present invention are described in detail above. It should be understood that a person skilled in the art can make many modifications and changes based on the concept of the present invention without creative work. Therefore, any technical solution that can be obtained by a person skilled in the art through logical analysis, reasoning or limited experiments based on the concept of the present invention on the basis of the prior art should be within the scope of protection determined by the claims.

Claims

1. A method for obtaining mechanical parameters of tunnel surrounding rock, characterized in that: The method comprises the following steps: S1: Select the rock mechanical parameters of the tunnel surrounding rock mass to be inverted and the calculation scheme of the construction parameter sample; S2: Perform numerical calculation on each constructed calculation scheme to obtain the displacement of the hole surrounding measurement points corresponding to each calculation scheme, and use the calculation scheme and the displacement of the corresponding hole surrounding measurement points as input parameter and output parameter sample sets; S3: Based on the input parameter and output parameter sample set obtained in step S2, a least squares support vector machine (LSSVM) numerical proxy model of the nonlinear mapping relationship between the surrounding rock mechanical parameters to be inverted and the displacement of the measuring points around the tunnel is established; S4: The improved firefly algorithm IFA is used to obtain the optimal combination of the hyperparameter regularization parameter γ and the Gaussian kernel parameter σ of the LSSVM numerical proxy model, and then the optimal IFA-LSSVM numerical proxy model is obtained; S5: Substitute the measured displacement data in the field into the optimal IFA-LSSVM numerical proxy model obtained in step S4 to obtain the final mechanical parameters of the rock and soil mass.

2. A method for obtaining mechanical parameters of tunnel surrounding rock according to claim 1, characterized in that: The specific steps of S1 are: S1-1: Select the mechanical parameters to be inverted by principal component analysis or factor sensitivity analysis; S1-2: Determine the value range of the rock mechanics parameters to be inverted for the tunnel surrounding rock mass based on the on-site investigation of the tunnel site; S1-3: Based on the value range, the Latin hypercube sampling method is used to construct a calculation scheme for parameter samples.

3. A method for obtaining mechanical parameters of tunnel surrounding rock according to claim 1, characterized in that: The specific steps of S4 are: S4-1: Construct an improved firefly algorithm; S4-2: Using the improved firefly algorithm of S4-1, the hyperparameter regularization parameter γ and Gaussian kernel parameter σ of the LSSVM numerical proxy model are optimized and calculated to obtain the optimal combination of the hyperparameter regularization parameter γ and the Gaussian kernel parameter σ. The optimal combination is substituted into the LSSVM numerical proxy model to obtain the optimal IFA-LSSVM numerical proxy model.

4. A method for obtaining mechanical parameters of tunnel surrounding rock according to claim 3, characterized in that: The steps of S4-2 are: Initialize the position of the firefly; Calculate the fitness function of the firefly according to the current position of the firefly; Update the position of the firefly. The updated position of the firefly is: x j+1 =β(t)x j +(1-β(t))x i +a(t)e j Among them, α(t) is the adaptive moving step size; is the attraction value; β0 is the attraction value at the light source (r=0); γ t is the adaptive light absorption coefficient; ε j is a random number vector that follows a uniform or normal distribution; r ij represents the Cartesian distance between any two firefly individuals i and j; Iterate the calculation until the number of iterations reaches the threshold, output the optimal combination of the hyperparameter regularization parameter γ and the Gaussian kernel parameter σ, substitute the optimal combination into the LSSVM numerical proxy model, and obtain the optimal IFA-LSSVM numerical proxy model.

5. A method for obtaining mechanical parameters of tunnel surrounding rock according to claim 4, characterized in that: The adaptive moving step size is: Among them, α0 is the initial moving step length, Itr max is the maximum number of iterations, c is an integer that determines the randomness decay rate, and t is the current number of iterations.

6. A method for obtaining mechanical parameters of tunnel surrounding rock according to claim 5, characterized in that: The adaptive light absorption coefficient is: Where γ0 is the initial light intensity absorption coefficient.

7. A method for obtaining mechanical parameters of tunnel surrounding rock according to claim 1, characterized in that: The calculation scheme for constructing parameter samples adopts the Latin hypercube sampling method.

8. The method for obtaining mechanical parameters of tunnel surrounding rock according to claim 1, characterized in that: Rock mechanical parameters include elastic modulus, Poisson's ratio, internal friction angle and cohesion.

9. A method for obtaining mechanical parameters of tunnel surrounding rock according to claim 8, characterized in that: Rock mechanical parameters are selected based on principal component analysis or factor sensitivity analysis methods.

10. The method for obtaining mechanical parameters of tunnel surrounding rock according to claim 1, characterized in that: The finite element method is used for numerical calculation.

Images