Advanced dynamic evaluation method for sanding characteristics of dolomite

Abstract

The invention discloses an advanced dynamic evaluation method for dolomite sanding characteristics. The advanced dynamic evaluation method comprises the following steps: determining mesoscopic and macroscopic thresholds of dolomite sanding degree grading as rock test static parameters based on an indoor test; extracting kinetic parameters and electrical parameters of the dolomite rock mass as advanced detection dynamic parameters based on geophysical advanced detection of missile-electricity combination; performing principal component analysis and correlation analysis on the rock test static parameters and advanced detection dynamic parameters, and constructing a dolomite sanding dynamic-static parameter quantitative conversion model by using RBFNN (Radial Basis Function Neural Network); training an intelligent grading evaluation model by taking the rock test static parameters as input and the sanding degree grade as output; and mapping the advanced detection dynamic parameters into rock test static parameters through the quantitative conversion model, inputting the mapped rock test static parameters into the intelligent grading evaluation model for verification and iterative optimization, and outputting a sanding degree grading result. Advanced prevention and control of sanding of the dolomite in front of the tunnel construction face can be achieved, and the ground disaster of sand collapse caused by sanding and sand gushing is avoided.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of geophysical exploration, and particularly relates to a dolomite sanding feature advanced dynamic evaluation method. BACKGROUND

[0002] Dolomite sanding is a special geological phenomenon that microcrystalline-fine crystalline structure dolomite is weathered into fine sand, gravel or block under the combined action of dissolution and weathering in a complex geological structure environment with multiple geological structure movements, and the quality of the rock mass is greatly reduced.

[0003] The tunnel surrounding rock deformation, working face instability, crown collapse and water inrush and sand phenomenon caused by the special karst phenomenon of dolomite sanding are more and more frequent. Dolomite sanding phenomenon is distributed all over the world, and there are dolomite sanding in southwest China such as Yunnan, Guizhou, Sichuan and Chongqing.

[0004] The stability of dolomite sanding tunnel surrounding rock is affected by its distribution space, sanding degree, and groundwater conditions. At present, it is difficult to accurately find out the distribution range of dolomite in front of the working face of the tunnel construction, the physical and mechanical properties and the groundwater occurrence environment. SUMMARY

[0005] The present application aims to overcome the problems of the prior art, and provides a dolomite sanding feature advanced dynamic evaluation method to solve the problems of unclear mechanism and mode cognition of dolomite sanding feature advanced dynamic evaluation and intelligent grading method, and to realize the prevention and control of dolomite sanding in front of the working face of the tunnel construction, and to avoid sanding and sand inrush collapse disasters.

[0006] To achieve the above technical purpose, the present application adopts the following technical scheme:

[0007] A dolomite sanding feature advanced dynamic evaluation method comprises the following steps:

[0008] Step S1, determining the micro and macro threshold values of dolomite sanding degree classification based on laboratory tests, and taking the determined micro and macro threshold values as rock test static parameters;

[0009] Step S2, extracting the dynamic parameters and electrical parameters of the dolomite rock mass based on the elastic-electric combined geophysical advanced detection, and taking the dynamic parameters and electrical parameters of the dolomite rock mass as advanced detection dynamic parameters;

[0010] Step S3, performing principal component analysis and correlation analysis on the rock test static parameters and the advanced detection dynamic parameters, and constructing a dolomite sanding dynamic-static parameter quantitative conversion model by using a radial basis function neural network (RBFNN) to output the mapping relationship from dynamic parameters to static parameters;

[0011] Step S4, training the intelligent grading evaluation model based on the autoencoder with the rock test static parameters as input and the sanding degree grade as output; mapping the advanced detection dynamic parameters to the rock test static parameters through the dolomite sanding dynamic-static parameter quantitative conversion model, inputting the mapped rock test static parameters into the intelligent grading evaluation model based on the autoencoder for verification and iterative optimization, and finally outputting the high-precision sanding degree grading result.

[0012] Further, the step S1 comprises:

[0013] Step S11, based on the micro-substance composition and structure and macroscopic characteristics of dolomite sanding, a dolomite sample library of different regions and different sanding degrees is established;

[0014] Step S12, using the dolomite sample library to carry out indoor tests, obtaining the macro-micro multi-property parameters of dolomite of different sanding degrees under the same test conditions, the indoor tests including rock physical and mechanical property tests and dolomite sanding tests under different stress paths;

[0015] Step S13, combining the micro-substance structure analysis of dolomite of different sanding degrees, the acoustic wave evolution law and deformation and failure mode of dolomite under different stress paths, analyzing the correlation between the macro-micro multi-property parameters of dolomite of different sanding degrees and the sanding degree grading, and determining the micro and macro thresholds of dolomite sanding degree grading as the rock test static parameters.

[0016] Further, the micro-substance composition of dolomite sanding includes rock mineral composition; the macroscopic characteristics of dolomite sanding include sanding degree, sanding strip thickness, sanding strip number, sanding strip characteristics and rock mass structure characteristics.

[0017] Further, the statistical analysis includes calculating the data distribution function, mean value, median, standard deviation and coefficient of variation of the physical property indoor test parameters of the rock sample, eliminating or repeatedly measuring abnormal data, and summarizing the distribution characteristics and differences of the physical property indoor test parameters of dolomite of different sanding degrees.

[0018] Further, the dynamic parameters include P-wave and S-wave velocities, density, dynamic elastic modulus and dynamic Poisson's ratio; the electrical parameters include dielectric constant, conductivity and resistivity.

[0019] Further, in the step S2:

[0020] The geophysical advanced detection based on the elastic-electric combination includes using a tunnel elastic wave advanced prediction instrument TEP, a multi-offset ground penetrating radar method MGPR and a transient electromagnetic method TEM.

[0021] The tunnel elastic wave advanced prediction instrument TEP is used to obtain the dynamic parameters of the dolomite rock mass;

[0022] The multi-offset ground penetrating radar method MGPR is used to obtain the dielectric constant and conductivity of the dolomite rock mass;

[0023] The transient electromagnetic method TEM is used to obtain the resistivity of the dolomite rock mass.

[0024] Further, in the step S2, the method for extracting dynamic parameters based on the elastic-electric combined geophysical advanced detection includes geophysical forward and full waveform inversion techniques; the geophysical forward includes elastic wave forward simulation and electromagnetic wave forward simulation, the elastic wave forward simulation is performed under a tunnel observation system based on time domain staggered grid finite difference through elastic wave wave equation, and the electromagnetic wave forward simulation is performed through finite difference by using a frequency domain electromagnetic wave equation under a transverse magnetic mode.

[0025] Further, the step S3 includes:

[0026] The dynamic parameters and electrical parameters of the dolomite rock mass extracted based on the elastic-electric combined geophysical advanced detection and the rock test static parameters determined through indoor tests are subjected to principal component analysis and correlation analysis, to obtain the correlation between the advanced detection dynamic parameters and the indoor test static parameters;

[0027] Based on the correlation between the advanced detection dynamic parameters and the indoor test static parameters, a dolomite sandification dynamic-static parameter quantitative conversion model is constructed by using a radial basis function neural network, multi-parameter correction of dolomite advanced evaluation is realized, a mapping relationship from dynamic parameters to static parameters is output, and the nonlinear degree of dolomite sandification degree intelligent grading and evaluation is reduced.

[0028] Further, the step S4 specifically includes:

[0029] Step S41, constructing an autoencoder intelligent grading and evaluation model: an autoencoder (AE) is used to construct an intelligent grading and evaluation model of the dolomite stratum tunnel advanced sandification degree;

[0030] Step S42, model training: rock test static parameters of different sandification degrees are taken as training data, sandification band thickness, sandification band number, sandification band characteristics, rock mass structure characteristics and other macro-graded bases are coded, and the sandification degree level is taken as a label together, and the autoencoder network parameter training is completed;

[0031] Step S43, basic accuracy verification: rock test static parameters of the training rock sample obtained through quantitative conversion of the advanced detection dynamic parameters are taken as the input of the autoencoder, and the grading accuracy of the autoencoder intelligent grading and evaluation model and the conversion accuracy of the dolomite sandification dynamic-static parameter quantitative conversion model are verified synchronously.

[0032] Step S44, generalization test: inputting the rock sample static parameters not participating in training and corresponding advanced detection dynamic parameters into the trained autoencoder intelligent grading evaluation model to verify the grading accuracy and generalization ability thereof;

[0033] Step S45, synergistic optimization: taking the grading precision of the autoencoder intelligent grading evaluation model and the conversion precision of the dolomite sanding dynamic-static parameter quantitative conversion model as constraint conditions, iteratively improving the precision of the dynamic-static parameter quantitative conversion model and the autoencoder intelligent grading evaluation model, and training and obtaining a high-accuracy and high-generalization tunnel dolomite sanding degree intelligent grading model;

[0034] Step S46, inputting the mapped rock test static parameters into the obtained tunnel dolomite sanding degree intelligent grading model in step S45, outputting high-precision sanding degree grading results, and realizing dynamic advanced evaluation of dolomite sanding stratum.

[0035] The present application has the following beneficial effects:

[0036] The present application provides a dolomite sanding feature advanced dynamic evaluation method, determines the micro and macro threshold values of dolomite sanding degree grading through indoor test and feature statistical analysis of dolomite static parameters with different sanding degrees; based on elastic-electric combined geophysical advanced detection dynamic parameter extraction, based on geophysical advanced detection numerical calculation, reveals the elastic wave and electromagnetic wave field propagation characteristics and spatio-temporal evolution law of dolomite sanding in tunnel space; constructs the quantitative conversion relationship of dolomite sanding dynamic-static parameters, realizes multi-parameter correction of dolomite advanced evaluation, weakens the nonlinearity degree of dolomite sanding degree intelligent grading and evaluation; can realize dynamic advanced evaluation of dolomite sanding stratum, effectively improves the advanced detection ability of dolomite sanding tunnel construction, improves the construction safety, and reduces the construction risk. BRIEF DESCRIPTION OF DRAWINGS

[0037] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings needed in the embodiments. Obviously, the drawings described below are only some embodiments of the present application, and other drawings can also be obtained by those skilled in the art without creative labor.

[0038] Figure 1 The dolomite sanding feature advanced dynamic evaluation and intelligent grading method flowchart provided for the embodiments of the present application;

[0039] Figure 2 The principal component analysis and correlation analysis process schematic diagram of advanced detection dynamic parameters and indoor test static parameters provided for the embodiments of the present application;

[0040] Figure 3 A training process schematic diagram of the dolomite sanding dynamic-static parameter quantitative conversion model based on the RBFNN provided by the embodiment of the present application is shown in the figure.

[0041] Figure 4 A training process schematic diagram of the intelligent grading evaluation model based on the auto-encoder provided by the embodiment of the present application is shown in the figure. DETAILED DESCRIPTION

[0042] To make the objectives, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. It should be understood that the drawings in the present application only serve the purpose of illustration and description, and are not used to limit the protection scope of the present application. The flowchart shows the operations implemented according to some embodiments of the present application. It should be understood that the operations of the flowchart can not be implemented in sequence, and the steps without logical context relationship can be reversed in sequence or implemented simultaneously. In addition, one or more other operations can be added to the flowchart or one or more operations can be removed from the flowchart under the guidance of the content of the present application.

[0043] As shown in the figure, the embodiment of the present application provides a dolomite sanding characteristic advanced dynamic evaluation method, which comprises: Figure 1

[0044] Step S1, determine the mesoscopic and macroscopic threshold values of dolomite sanding degree grading based on indoor test, and take the determined mesoscopic and macroscopic threshold values as rock test static parameters;

[0045] Specifically, step S1 comprises:

[0046] Step S11, based on the dolomite sanding mesoscopic material composition and structure and macroscopic characteristics, establish a dolomite sample library of different regions and different sanding degrees;

[0047] Specifically, to establish a dolomite sample library, dolomite rock samples of different sanding degrees need to be collected, and the rock samples are described and grouped according to the commonly used dolomite sanding qualitative grading standard (Table 1) at the present stage.

[0048] Table 1 Macroscopic grading standard of dolomite sanding degree

[0049]

[0050] The dolomite sanding mesoscopic material composition comprises rock mass mineral composition; and the dolomite sanding macroscopic characteristics comprise sanding degree, sanding band thickness, sanding band number, sanding band characteristics and rock mass structure characteristics.

[0051] ​In the process of collecting dolomite samples with different degrees of sanding, preferably, in order to reduce the difference between the state of the rock sample after drilling and coring and the state of the rock sample in situ in the tunnel, after the rock sample is taken out on the spot, the rock sample is packaged with the advanced ePTFE film in the current material field. The material has the advantages of light weight (negligible), water resistance, acid and alkali corrosion resistance, non-combustibility, and extreme stability to oxygen and ultraviolet light, which can effectively reduce the change of the state of the rock sample during transportation, storage and testing.

[0052] Preferably, for the loose and disintegrated dolomite rock samples with high sanding degree and water content, a cubic test tank is made, and the test tank material is metal or wood according to the needs of test parameters, so as to meet the needs of different sanding degree rock samples and different measurement parameter test methods.

[0053] Step S12, indoor test is carried out by using the dolomite sample library, the indoor test parameters of each physical property of the rock sample obtained under the same test conditions are statistically analyzed, and the macroscopic and microscopic multi-property parameters of dolomite with different sanding degrees are obtained. The indoor test includes rock physical and mechanical property test and dolomite sanding test under different stress paths; the statistical analysis includes calculating the data distribution function, mean value, median, standard deviation and coefficient of variation of the indoor test parameters of each physical property of the rock sample, removing or repeatedly measuring the abnormal data, and summarizing the distribution characteristics and differences of the indoor test parameters of each physical property of dolomite with different sanding degrees.

[0054] Step S13, the correlation between the macroscopic and microscopic multi-property parameters of dolomite with different sanding degrees and the sanding degree classification is analyzed by combining the microscopic material structure analysis of dolomite with different sanding degrees, the acoustic wave evolution law and deformation and failure mode of dolomite under different stress paths, and the microscopic and macroscopic threshold values of dolomite sanding degree classification are determined as the rock test static parameters.

[0055] Step S2, the dynamic parameters and electrical parameters of the dolomite rock mass are extracted based on the elastic-electric combined geophysical advanced detection, and the dynamic parameters and electrical parameters of the dolomite rock mass are used as the advanced detection dynamic parameters; the dynamic parameters include longitudinal and transverse wave velocity, density, dynamic elastic modulus and dynamic Poisson's ratio, and the electrical parameters include dielectric constant, conductivity and resistivity.

[0056] The elastic-electric combined geophysical advanced detection method includes using a tunnel elastic wave advanced prediction instrument TEP, a multi-offset ground penetrating radar method MGPR and a transient electromagnetic method TEM. The tunnel elastic wave advanced prediction instrument TEP is used to obtain the dynamic parameters of the dolomite rock mass; the multi-offset ground penetrating radar method MGPR is used to obtain the dielectric constant and conductivity of the dolomite rock mass; and the transient electromagnetic method TEM is used to obtain the resistivity of the dolomite rock mass.

[0057] Specifically, the density of the dolomite sanding rock sample can be tested by the wax sealing method, and the natural density of the rock sample can be represented as:

[0058] p 0 e = m 0 / [ ( m 1 − m 2 ) / p wt − ( m 1 − m 0 ) / p n ] ,

[0059] wherein represents the test natural density, m0 represents the natural mass of the test piece, m1 represents the mass of the wax-sealed test piece, m2 represents the water weighing of the wax-sealed test piece, and p wt represents the density of water, and p n represents the density of wax.

[0060] The longitudinal and transverse wave velocities of dolomite with different sanding degrees are obtained by using longitudinal and transverse wave transducers through direct wave method (i.e., direct penetration method), and the relationship between the longitudinal and transverse wave velocities of the rock sample is as follows:

[0061]

[0062]

[0063] wherein represents the test longitudinal wave velocity, represents the test transverse wave velocity, and l represents the distance between the center points of the transmitting and receiving transducers, t p represents the propagation time of the longitudinal wave in the rock sample, t s represents the propagation time of the transverse wave in the rock sample, and t0 represents the zero delay of the instrument system.

[0064] The dielectric constant ε of the rock sample is r determined by the capacitance method:

[0065]

[0066] wherein represents the test dielectric constant, C0 represents the vacuum capacitance between the two plates, and C r represents the capacitance after adding the rock sample.

[0067] The resistivity of the rock sample is determined by the four-pole method:

[0068]

[0069] wherein R e represents the test resistivity, S represents the cross-sectional area of the rock sample, L is the electrode spacing, U represents the voltage, and I is the current.

[0070] wherein the extraction method of advanced detection dynamic parameters based on the elastic-electric combined geophysical advanced detection includes geophysical forward and full waveform inversion techniques.

[0071] The geophysical forward includes elastic wave forward simulation and electromagnetic wave forward simulation:

[0072] The elastic wave forward simulation is based on time domain staggered grid finite difference under tunnel observation system, by elastic wave wave equation, wherein the elastic wave two-dimensional first-order velocity-stress equation can be expressed as:

[0073]

[0074] Wherein, in two-dimensional x-z space, (vx, vz) represents velocity component, (σx, σz) represents stress component, L represents Lamé constant of elastic unit body and satisfies the following relationship:

[0075]

[0076] Wherein v p And v s Respectively represent the velocity of longitudinal wave and transverse wave, and ρ represents density.

[0077] The electromagnetic wave forward simulation adopts finite difference of frequency domain electromagnetic wave equation under transverse magnetic (TM) mode:

[0078]

[0079] Wherein x, y, z respectively represent three spatial directions, E and H respectively represent electric field intensity and magnetic field intensity, ε e And η represent equivalent permittivity containing permittivity and conductivity information, and magnetic permeability coefficient.

[0080] The full waveform inversion can include the following steps S21-S24 described in the following steps:

[0081] S21, construct the objective function with L2 norm of observation data and simulation data:

[0082] J ( m ) = 1 2 ∑ s , r [ d syn ( s , r , m ) − d obs ( s , r ) ] 2 + g ‖ m − m 0 ‖ 2 2 ,

[0083] Wherein J represents the objective function, d syn And d obs Respectively represent simulation data and observation data, s, r, m respectively represent field source variable, receiving point variable and model parameter to be solved, and γ represents regularization term weight. The mapping from data space to model space is gradient, and the derivative of the objective function to the required inversion model parameter can be obtained:

[0084] ∂ J ∂ m = ∑ s , r ∂ [ d syn ( m ) − d obs ] ∂ m ( d d ) = ∑ s , r ∂ d ( m ) ∂ m ( d d ) = d m ,

[0085] ​According to the adjoint state method and the constitutive relation between parameters, the gradients of the P-wave velocity, S-wave velocity and density can be expressed as:

[0086]

[0087] The gradients of the permittivity and conductivity can be expressed as:

[0088]

[0089] Where T represents the back-propagated wavefield.

[0090] S22, by designing multiple tunnel dolomite sanding geological models, obtaining the optimal inversion strategy of multi-parameter FWI, improving the convergence of inversion while suppressing crosstalk noise, and improving the accuracy of the inversion result. The advanced dynamic Poisson's ratio μ d and dynamic Young's modulus E d can be calculated by the following formula:

[0091] m d = ( v p v s ) 2 − 2 2 [( v p v s ) 2 − 1 ] ,

[0092]

[0093] S23, a mixed field source coding strategy of tunnel TEP and MGPR data is adopted to meet the timeliness requirement of tunnel surrounding rock advanced evaluation, and the calculation time of FWI is shortened. The single field source data obtained by multiple field source excitation is combined into a mixed field source in the form of coding, and a mixed field source data can be expressed as:

[0094] S = ∑ i = 1 Ns A rand · PH rand [ s i ( x , z , t ) * d ( t − A t rand ) ] ，

[0095] Where S represents a mixed field source data, A rand represents a random amplitude polarity coding, si represents a function at the field source position (x, z), δ (t-Δt rand ) represents a δ function with a random delay time of Δt rand , · represents multiplication, * represents convolution, PH rand (.) represents a random phase coding operator, which can be expressed as:

[0096]

[0097] Where θ rand represents a random phase rotation angle, and Hi (α) represents the Hilbert transform of signal α.

[0098] Further, in the inversion iteration, the iteration number k is set to an integer multiple of 10, and the observation data is re-encoded to ensure the convergence of the inversion. The feasibility of mixed field source encoding under the tunnel observation system and the effectiveness of the encoding method are obtained, and the best encoding strategy to ensure the convergence and calculation efficiency of FWI is summarized.

[0099] S24, the TEM observation system is arranged at the working face and the floor respectively, the forward simulation of the tunnel space magnetic field strength is carried out in the higher precision observation mode for the dolomite sanding area, and the TEM total field integral equation is:

[0100]

[0101] In the formula: and are the Green functions of the primary electric field and the secondary electric field respectively, is the stiffness matrix obtained by using the internal edge function for the weight vector and the field vector, is the stiffness matrix obtained by using the internal edge function for the weight vector and the boundary edge function for the field vector, and E is the conduction current density.

[0102] After the primary electric field strength is obtained, the secondary magnetic field strength can be obtained according to the following formula, that is:

[0103]

[0104] In the formula: is the primary magnetic field strength, is the Green function of the secondary magnetic field, is the conductivity,

[0105] The least square inversion method of tunnel TEM data is studied, and the objective function is constructed under the mixed norm in the data space and the model space:

[0106]

[0107] Wherein, C is the smoothness matrix. Derive the objective function with respect to the model parameters, and update the initial model by local optimization algorithm iteration to obtain the high-precision resistivity distribution of dolomite sanding tunnel.

[0108] Step S3, the rock test static parameters and the advanced detection dynamic parameters are subjected to principal component analysis and correlation analysis, a dolomite sanding dynamic-static parameter quantitative conversion model is constructed by using a radial basis function neural network (RBFNN), and a mapping relationship quantitative conversion model of dynamic parameters to static parameters is output;

[0109] The step S3 specifically includes:

[0110] ​S31, respectively, statistics of different sanding degree dolomite advanced detection dynamic parameters and rock test static parameters of the distribution characteristics, and the principle component analysis (Principle Component Analysis, PCA), the generation of comprehensive data component reduces the correlation between the original data, weaken the complexity of data analysis.

[0111] S32, the rock sample is divided into groups, the dynamic parameters and static parameters of each rock sample after PCA comprehensive data component is based on correlation coefficient correlation analysis (Correlation Analysis, CA):

[0112]

[0113] Wherein cov represents the covariance, s x and s y respectively represent the standard deviation, X, Y are the dynamic parameters and static parameters of the rock sample, is the mean of the dynamic parameters, is the mean of the static parameters, n is the sample group number, i is the sample serial number

[0114]

[0115] S33, multiple replacement rock sample grouping, analysis of the correlation between the advanced detection dynamic parameters and the indoor test static parameters, advanced detection dynamic parameters and indoor test static parameters PCA and CA analysis process as Figure 2 shown.

[0116] S34, based on the correlation between the advanced detection dynamic parameters and the indoor test static parameters, the radial basis function neural network RBFNN is used to establish the dolomite sanding dynamic-static parameter quantitative conversion model:

[0117]

[0118] Wherein x and y represent input and output respectively, w represents the network weight to be trained, and c represents the center of the radial basis function.

[0119] S35, the advanced detection dynamic parameters of each sanding degree are taken as input, and the corresponding indoor test static parameters are taken as output, the RBFNN is trained, and the data not participating in the training is used as the test set to test the accuracy and generalization of the prediction value of the constructed dolomite sanding dynamic-static parameter quantitative conversion model. The training process of the quantitative conversion model based on RBFNN is shown in Figure 3 .

[0120] Step S4, training the intelligent grading evaluation model based on the autoencoder with the rock test static parameters as input and the sanding degree grade as output; mapping the advanced detection dynamic parameters to the rock test static parameters through the dolomite sanding dynamic-static parameter quantitative conversion model, inputting the intelligent grading evaluation model based on the autoencoder for verification and iterative optimization, and finally outputting the high-precision sanding degree grading result.

[0121] The step S4 specifically comprises:

[0122] Step S41, constructing an intelligent grading evaluation model of the autoencoder: an intelligent grading evaluation model of the dolomite stratum tunnel advanced sanding degree is constructed by using an autoencoder (AE).

[0123] Step S42, model training: the rock test static parameters of different sanding degrees are taken as training data, the sanding strip thickness, sanding strip number, sanding strip characteristics, rock mass structure characteristics and other macro-graded basis are coded, and the sanding degree grade is taken as a label together, the autoencoder (AE) network parameter training is completed, and the training process is as shown in Figure 4 .

[0124] Step S43, basic precision verification: the rock test static parameters obtained by quantitatively converting the advanced detection dynamic parameters of the training rock sample are taken as the input of the autoencoder (AE), and the grading accuracy of the autoencoder intelligent grading evaluation model and the conversion accuracy of the dolomite sanding dynamic-static parameter quantitative conversion model are verified synchronously.

[0125] Step S44, generalization test: the static parameters of the rock sample not participating in the training and the corresponding advanced detection dynamic parameters are input into the trained autoencoder intelligent grading evaluation model to verify the grading accuracy and generalization ability thereof.

[0126] Step S45, collaborative optimization: taking the grading accuracy of the autoencoder intelligent grading evaluation model and the conversion accuracy of the dolomite sanding dynamic-static parameter quantitative conversion model as constraint conditions, the precision of the dynamic-static parameter quantitative conversion model and the autoencoder intelligent grading evaluation model is iteratively improved, a high-accuracy and high-generalization tunnel dolomite sanding degree intelligent grading model is trained and obtained.

[0127] Step S46, inputting the mapped rock test static parameters into the tunnel dolomite sanding degree intelligent grading model obtained in step S45, outputting the high-precision sanding degree grading result, and realizing the dynamic advanced evaluation of the dolomite sanding stratum.

[0128] The above merely illustrates the specific embodiments of the present application, but the protection scope of the present application is not limited thereto, any changes or replacements within the technical scope disclosed by the present application, which can be easily thought by any person skilled in the art, should be covered within the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A method for advanced dynamic evaluation of dolomite sandification characteristics, characterized in that: include: Step S1: determining the microscopic and macroscopic thresholds for the sandification degree classification of dolomite based on indoor experiments, and using the determined microscopic and macroscopic thresholds as static parameters of rock tests; Step S2: extracting the dynamic parameters and electrical parameters of the dolomite rock mass based on the combined elastic-electrical geophysical advance detection, and using the dynamic parameters and electrical parameters of the dolomite rock mass as the advanced detection dynamic parameters; Step S3: performing principal component analysis and correlation analysis on the static parameters of the rock test and the dynamic parameters of the advanced detection, and constructing a quantitative conversion model of the dynamic-static parameters of dolomite sandification using a radial basis function neural network (RBFNN) to output a mapping relationship from the dynamic parameters to the static parameters; Step S4: Using the static parameters of the rock test as input and the sandification degree grade as output, train an intelligent grading evaluation model based on an autoencoder; map the advanced detection dynamic parameters to the rock test static parameters through the dolomite sandification dynamic-static parameter quantitative conversion model; input the mapped rock test static parameters into the intelligent grading evaluation model based on the autoencoder for verification and iterative optimization, and finally output a high-precision sandification degree classification result.

2. The method for advanced dynamic evaluation of dolomite sandification characteristics according to claim 1, characterized in that: The step S1 comprises: Step S11: establishing a dolomite sample library with different regions and different degrees of sandification based on the microscopic material composition and structure and macroscopic characteristics of dolomite sandification; Step S12: Conducting indoor tests using the dolomite sample library to statistically analyze the physical property parameters of rock samples obtained under the same test conditions for dolomites with different degrees of sandification, thereby obtaining macroscopic and microscopic multi-physical property parameters of dolomites with different degrees of sandification. The indoor tests include tests on rock physical and mechanical properties and tests on dolomite sandification with different stress paths. Step S13: Combined with the microscopic material structure analysis of dolomites with different sandification degrees, the acoustic wave evolution law and deformation and failure mode of dolomites under different stress paths, the correlation between the macroscopic and microscopic multi-physical parameters of dolomites with different sandification degrees and the sandification degree classification is analyzed, and the microscopic and macroscopic thresholds of the dolomite sandification degree classification are determined as the static parameters of the rock test.

3. The method for advanced dynamic evaluation of dolomite sandification characteristics according to claim 2, characterized in that: The microscopic material composition of the dolomite sandification includes the mineral components of the rock mass; the macroscopic characteristics of the dolomite sandification include the degree of sandification, the thickness of the sandification bands, the number of sandification bands, the characteristics of the sandification bands and the rock mass structure characteristics.

4. The method for advanced dynamic evaluation of dolomite sandification characteristics according to claim 2, characterized in that: The statistical analysis includes calculating the data distribution function, mean value, median, standard deviation and coefficient of variation of the indoor test parameters of various physical properties of rock samples, eliminating abnormal data or repeating the measurement, and summarizing the distribution characteristics and differences of the indoor test parameters of various physical properties of dolostones with different degrees of sandification.

5. The method for advanced dynamic evaluation of dolomite sandification characteristics according to claim 1, characterized in that: The dynamic parameters include longitudinal and transverse wave velocities, density, dynamic elastic modulus, and dynamic Poisson's ratio; the electrical parameters include dielectric constant, conductivity, and resistivity.

6. A method for advanced dynamic evaluation of dolomite sandification characteristics according to claim 5, characterized in that: In the step S2: The geophysical advance detection based on the combination of elasticity and electricity includes the use of tunnel elastic wave advance predictor TEP, multi-offset ground penetrating radar method MGPR and transient electromagnetic method TEM; The tunnel elastic wave advance predictor TEP is used to obtain the dynamic parameters of the dolomite rock mass; The multi-offset ground penetrating radar method MGPR is used to obtain the dielectric constant and conductivity of the dolomite rock mass; The transient electromagnetic method (TEM) is used to obtain the resistivity of the dolomite rock mass.

7. The method for advanced dynamic evaluation of dolomite sandification characteristics according to claim 5, characterized in that: In the step S2: a method for extracting dynamic parameters of geophysical advance detection based on elastic-electromechanical joint advance detection includes geophysical forward modeling and full waveform inversion technology; the geophysical forward modeling includes elastic wave forward simulation and electromagnetic wave forward simulation, the elastic wave forward simulation is performed through the elastic wave wave equation in a tunnel observation system based on time domain staggered grid finite difference, and the electromagnetic wave forward simulation uses the frequency domain electromagnetic wave equation in the transverse magnetic mode for finite difference.

8. The method for advanced dynamic evaluation of dolomite sandification characteristics according to claim 1, characterized in that: The step S3 comprises: Principal component analysis and correlation analysis were performed on the dynamic and electrical parameters of the dolomite rock mass extracted by combined elastic-electrical geophysical advance detection and the static parameters of the rock test determined by indoor tests to obtain the correlation between the dynamic parameters of the advance detection and the static parameters of the indoor tests. Based on the correlation between the dynamic parameters of advance detection and the static parameters of indoor tests, a radial basis function neural network is used to construct a quantitative conversion model of dynamic-static parameters of dolomite sandification. This realizes multi-parameter correction for dolomite advance evaluation and outputs the mapping relationship from dynamic parameters to static parameters, thereby reducing the nonlinearity of the intelligent classification and evaluation of dolomite sandification degree.

9. The method for advanced dynamic evaluation of dolomite sandification characteristics according to claim 1, characterized in that: The step S4 specifically includes: Step S41, constructing an autoencoder intelligent grading evaluation model: using an autoencoder (AE) to construct an intelligent grading evaluation model for the advanced sandification degree of a dolomite stratum tunnel; Step S42, model training: using static parameters of rock tests with different degrees of sandification as training data, encoding macroscopic classification criteria such as sandification band thickness, number of sandification bands, sandification band characteristics, and rock mass structural characteristics, and using these together with the sandification degree level as labels to complete the autoencoder network parameter training; Step S43, basic accuracy verification: using the static parameters of the rock test obtained by quantitatively converting the dynamic parameters of the training rock sample through advance detection as the input of the autoencoder, and simultaneously verifying the classification accuracy of the autoencoder intelligent classification evaluation model and the conversion accuracy of the dolomite sandification dynamic-static parameter quantitative conversion model; Step S44, generalization test: input the static parameters of the rock samples that have not participated in the training and the corresponding advanced detection dynamic parameters into the trained autoencoder intelligent grading evaluation model to verify its grading accuracy and generalization ability; Step S45, collaborative optimization: using the classification accuracy of the autoencoder intelligent classification evaluation model and the conversion accuracy of the dynamic-static parameter quantitative conversion model of dolomite sandification as mutual constraints, iteratively improve the accuracy of the dynamic-static parameter quantitative conversion model and the autoencoder intelligent classification evaluation model, and train and obtain a highly accurate and generalizable intelligent classification model for tunnel dolomite sandification degree; Step S46: Input the mapped rock test static parameters into the tunnel dolomite sandification degree intelligent classification model obtained in step S45, output a high-precision sandification degree classification result, and realize dynamic advanced evaluation of dolomite sandification strata.

Images