Multi-parameter rockburst prediction method and system based on machine learning

Abstract

The invention provides a multi-parameter rockburst prediction method and system based on machine learning, and relates to the technical field of deep tunnel construction. The method comprises the steps that crustal stress information of a construction area is measured through a hydraulic fracturing method, and tangential crustal stress of a tunnel is determined; in the area where the rockburst tendency needs to be evaluated, on-site rocks are collected, and a rock sample is prepared; performing a uniaxial compression test on the test piece to obtain the uniaxial compressive strength of the test piece; performing a Brazilian disc splitting test on the test piece to obtain the tensile strength of the test piece; calculating a crustal stress coefficient, a brittleness coefficient and an elastic energy index of the surrounding rock according to the measured tangential crustal stress, uniaxial compressive strength and tensile strength; collecting existing rockburst data, and establishing a rockburst case database; training a plurality of machine learning models by using the rockburst case database, and establishing a rockburst strength prediction model; and inputting the crustal stress and the surrounding rock parameters into the rockburst strength prediction model, and predicting the rockburst strength.

Description

Technical Field

[0001] The present invention belongs to the technical field of deep tunnel construction, and relates to a multi-parameter rockburst prediction method and system based on machine learning. Background Art

[0002] With the exploitation of deep resources and the construction of deep buried tunnels in my country, the excavation of high-stress rock mass is increasingly involved. Deep rock mass stores a large amount of strain energy under high-stress conditions. The unloading of tunnel excavation breaks the original balance of the surrounding rock, and the strain energy in the rock mass is released sharply, driving the dynamic fracture of the rock mass, and then rock burst occurs. Rock burst is characterized by suddenness and violence, which is specifically manifested in the explosion, loosening, spalling, and ejection of rock blocks, causing serious damage to construction personnel and equipment, and seriously affecting project safety and construction progress.

[0003] In order to effectively predict rockbursts, domestic and foreign scholars have conducted extensive research on this issue in recent decades and proposed a large number of rockburst occurrence criteria, such as Russense criterion, Barton criterion, Hoek criterion, Tao Zhenyu criterion, etc. These rockburst occurrence criteria predict the degree of rockburst damage by comparing rock mechanical properties with ground stress conditions. The main rock mechanical properties include uniaxial compressive strength, tensile strength, brittleness index, elastic energy coefficient, etc.

[0004] However, the existing rockburst criteria are mostly based on a single rock mechanical property index, which will bring errors to the rockburst prediction results. When using multiple rock mechanical property indexes for rockburst prediction, the correlation between the mechanical property indexes and the weight of their influence on the prediction results are difficult to determine. Summary of the invention

[0005] The purpose of the present invention is to provide a multi-parameter rockburst prediction method and system based on machine learning to solve the problem that the rockburst criterion in the prior art is mostly based on a single rock mechanical property index, which will cause errors in the rockburst prediction results. When using multiple rock mechanical property indicators for rockburst prediction, the correlation between the mechanical property indicators and the weight of the impact on the prediction result are difficult to determine.

[0006] In order to achieve the above object, the present invention adopts the following technical solutions:

[0007] A multi-parameter rockburst prediction method based on machine learning includes the following steps:

[0008] Obtain the tangential ground stress in the construction area and the uniaxial compressive strength and tensile strength of rock samples in the construction area;

[0009] Calculate the geostress coefficient, brittleness coefficient and elastic energy index of the surrounding rock based on the tangential geostress, uniaxial compressive strength and tensile strength;

[0010] Based on the existing rockburst data, a rockburst case database is established, and the machine learning model is trained based on the rockburst case database to establish a rockburst intensity prediction model;

[0011] The tangential ground stress, ground stress coefficient of surrounding rock, brittleness coefficient and elastic energy index are input into the rock burst strength prediction model to predict the rock burst strength.

[0012] Preferably, the tangential geostress is determined by measuring the geostress information of the construction area through hydraulic fracturing to clarify the tangential geostress of the tunnel, which specifically includes:

[0013] Arrange geostress measuring points according to the geological structure of the deep tunnel construction area, including the number, location and hole depth of the measuring points;

[0014] The hydraulic fracturing method is used to determine the geostress state of the deep tunnel construction area. By drilling holes at the rock mass measuring point, sealing the two ends of the test rock layer with a packer, injecting high-pressure liquid into the sealed borehole, pressurizing until the hole wall breaks, and recording the pressure change over time. The geostress state of the deep tunnel construction area is calculated using the corresponding formula based on the fracture pressure, pump shut-off pressure and fracture orientation, and the tangential geostress σ of the tunnel is determined. θ .

[0015] Preferably, the rock samples in the construction area are sampled by drilling using a coring machine, and the excavation speed of the coring machine is controlled at 0.5mm / min to 2mm / min; after drilling and coring, 4 cylindrical rock samples with a diameter of 50mm and a height of 100mm and 3 disc rock samples with a diameter of 50mm and a thickness of 25mm are prepared using the cores.

[0016] Preferably, obtaining the uniaxial compressive strength of the rock sample comprises:

[0017] The cylindrical rock sample was subjected to axial displacement loading at a loading rate of 0.02 mm / s using a compression testing machine until the sample was damaged, and the uniaxial compressive strength σ of the sample under uniaxial compression conditions was obtained. c and axial stress-strain curves;

[0018] Repeat the test three times, and take the average value of the uniaxial compressive strength measured by the three groups of tests as the uniaxial compressive strength σ of the rock. c .

[0019] Preferably, the tensile strength of the rock sample is obtained by a Brazilian disc splitting test, which specifically includes: radially loading the disc-shaped rock sample at a loading rate of 0.3-0.5 MPa / s using a compression testing machine until the sample is damaged, and recording the maximum damage load P;

[0020] Formula σ for calculating rock tensile strength using Brazilian disk splitting test t =2P / (πdT), for rock tensile strength σ t Calculate, where d is the diameter of the disk and T is the thickness of the disk;

[0021] Repeat the test three times, and take the average value of the tensile strength measured in the three groups of tests as the tensile strength σ of the rock. t .

[0022] Preferably, the calculation of the in-situ stress coefficient, brittleness coefficient and elastic energy index of the surrounding rock according to the measured tangential in-situ stress, uniaxial compressive strength and tensile strength comprises:

[0023] According to the calculation formula of surrounding rock stress coefficient S, S = σ θ / σ c , calculate the geostress coefficient of surrounding rock;

[0024] According to the calculation formula of surrounding rock brittleness coefficient B, B = σ c / σ t , calculate the brittleness coefficient of surrounding rock;

[0025] The cylindrical rock specimens were subjected to axial displacement loading at a loading rate of 0.02 mm / s using a compression testing machine until the load reached 0.8 times σ c Then, the sample was axially unloaded at an unloading rate of 0.02 mm / s until the axial load was 0 MPa, and the axial stress-strain curve of the sample was recorded during the whole process;

[0026] Integrate the area enclosed under the unloading curve to obtain the elastic strain energy E e ;

[0027] The area enclosed under the loading curve is integrated to obtain the total energy of the loading process, and then the plastic strain energy E is obtained by subtracting the elastic strain energy from the total energy. p ;

[0028] According to the calculation formula of surrounding rock elastic strain energy index Wet, Wet=E e / E p , calculate the elastic strain energy index of surrounding rock.

[0029] Preferably, the training of the machine learning model according to the rockburst case database to establish the rockburst intensity prediction model specifically includes:

[0030] The tangential in-situ stress, uniaxial compressive strength, tensile strength, in-situ stress coefficient, brittleness coefficient, and elastic energy index of the surrounding rock in the rockburst case database are used as input variables, and the rockburst intensity is used as the output variable;

[0031] Using the rockburst case database, machine learning models such as decision tree model, random forest model, support vector machine model, neural network model, Gaussian process regression model, and generalized additive model are trained to establish a rockburst intensity prediction model.

[0032] Preferably, the step of inputting the tangential ground stress, the ground stress coefficient, the brittleness coefficient and the elastic energy index of the surrounding rock into the rockburst intensity prediction model to predict the rockburst intensity specifically includes:

[0033] The tangential ground stress of the rockburst test area, the uniaxial compressive strength, tensile strength, ground stress coefficient, brittleness coefficient, and elastic energy index of the surrounding rock are input into the machine learning model;

[0034] The decision tree model, random forest model, support vector machine model, neural network model, Gaussian process regression model, generalized additive model and other machine learning models were used to predict the rockburst intensity respectively. The rockburst intensities predicted by multiple machine learning models were compared, and the prediction result with the most serious damage was taken as the final rockburst intensity prediction value.

[0035] A multi-parameter rockburst prediction system based on machine learning, comprising:

[0036] An acquisition unit is used to acquire the tangential ground stress in the construction area and the uniaxial compressive strength and tensile strength of the rock samples in the construction area;

[0037] A calculation unit is used to calculate the in-situ stress coefficient, brittleness coefficient and elastic energy index of the surrounding rock according to the tangential in-situ stress, uniaxial compressive strength and tensile strength;

[0038] A model training unit is used to establish a rock burst case database based on existing rock burst data, and to train a machine learning model based on the rock burst case database to establish a rock burst intensity prediction model;

[0039] The prediction unit is used to input the tangential ground stress, the ground stress coefficient of the surrounding rock, the brittleness coefficient and the elastic energy index into the rock burst strength prediction model to predict the rock burst strength.

[0040] Preferably, the calculation of the in-situ stress coefficient, brittleness coefficient and elastic energy index of the surrounding rock according to the measured tangential in-situ stress, uniaxial compressive strength and tensile strength comprises:

[0041] According to the calculation formula of surrounding rock stress coefficient S, S = σ θ / σ c , calculate the geostress coefficient of surrounding rock;

[0042] According to the calculation formula of surrounding rock brittleness coefficient B, B = σ c / σ t , calculate the brittleness coefficient of surrounding rock;

[0043] The cylindrical rock specimens were subjected to axial displacement loading at a loading rate of 0.02 mm / s using a compression testing machine until the load reached 0.8 times σ c Then, the sample was axially unloaded at an unloading rate of 0.02 mm / s until the axial load was 0 MPa, and the axial stress-strain curve of the sample was recorded during the whole process;

[0044] Integrate the area enclosed under the unloading curve to obtain the elastic strain energy E e ;

[0045] The area enclosed under the loading curve is integrated to obtain the total energy of the loading process, and then the plastic strain energy E is obtained by subtracting the elastic strain energy from the total energy. p ;

[0046] According to the calculation formula of surrounding rock elastic strain energy index Wet, Wet=E e / E p , calculate the elastic strain energy index of surrounding rock.

[0047] Compared with the prior art, the present invention has the following beneficial effects:

[0048] 1) This method comprehensively considers multiple rock mechanical properties such as uniaxial compressive strength, tensile strength, brittleness index, elastic energy coefficient, etc. Compared with the existing rock burst criterion based on a single rock mechanical property indicator, this method has higher accuracy.

[0049] 2) A large number of rockburst cases are used to train the machine learning model, and then a rockburst intensity prediction model is established. Therefore, this method will have good applicability to different geological conditions.

[0050] 3) Using this method to predict rock bursts during the construction of deep tunnels can reduce engineering risks, improve economic benefits, and ensure the safety of construction personnel and equipment to a certain extent, and has good application value. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments are briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without creative work.

[0052] Figure 1 is a flow chart of the method of the present invention;

[0053] Figure 2 is a flow chart of a method according to an embodiment of the present invention;

[0054] Figure 3 It is a schematic diagram of elastic strain energy coefficient calculation of the present invention;

[0055] Figure 4 The figure is a flow chart of the execution of the rockburst intensity prediction model of the present invention. DETAILED DESCRIPTION

[0056] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Generally, the components of the embodiments of the present invention described and shown in the drawings here can be arranged and designed in various different configurations.

[0057] Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the invention claimed for protection, but merely represents selected embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0058] It should be noted that similar reference numerals and letters denote similar items in the following drawings, and therefore, once an item is defined in one drawing, further definition and explanation thereof is not required in subsequent drawings.

[0059] In the description of the embodiments of the present invention, it should be noted that if the terms "upper", "lower", "horizontal", "inner", etc. indicate an orientation or positional relationship based on the orientation or positional relationship shown in the drawings, or the orientation or positional relationship in which the product of the invention is usually placed when in use, it is only for the convenience of describing the present invention and simplifying the description, and does not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation on the present invention. In addition, the terms "first", "second", etc. are only used to distinguish the description, and cannot be understood as indicating or implying relative importance.

[0060] In addition, if the term "horizontal" appears, it does not mean that the component must be absolutely horizontal, but can be slightly tilted. For example, "horizontal" only means that its direction is more horizontal than "vertical", which does not mean that the structure must be completely horizontal, but can be slightly tilted.

[0061] In the description of the embodiments of the present invention, it is also necessary to explain that, unless otherwise clearly specified and limited, the terms "set", "install", "connect", and "connect" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be indirectly connected through an intermediate medium, or it can be the internal connection of two components. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0062] The present invention is further described in detail below in conjunction with the accompanying drawings:

[0063] With the development of computer technology, machine learning provides a solution for predicting results based on multiple parameters. By using existing rock burst cases, taking ground stress information and rock uniaxial compressive strength, tensile strength, brittleness index, elastic energy coefficient and other parameters as input variables, and taking rock burst intensity as output variable, the machine learning model is trained and the influence of multiple parameters is comprehensively considered, which is expected to make more accurate predictions of rock bursts. Figure 1 The present invention discloses a multi-parameter rockburst prediction method based on machine learning, comprising the following steps:

[0064] S1: Obtain the tangential ground stress in the construction area and the uniaxial compressive strength and tensile strength of the rock samples in the construction area;

[0065] S2: Calculate the in-situ stress coefficient, brittleness coefficient and elastic energy index of the surrounding rock based on the tangential in-situ stress, uniaxial compressive strength and tensile strength;

[0066] S3: Based on the existing rockburst data, a rockburst case database is established, and the machine learning model is trained based on the rockburst case database to establish a rockburst intensity prediction model;

[0067] S4: The tangential ground stress, ground stress coefficient, brittleness coefficient and elastic energy index of the surrounding rock are input into the rock burst strength prediction model to predict the rock burst strength.

[0068] In some embodiments, see Figure 2 , a multi-parameter rockburst prediction method based on machine learning, including:

[0069] S1: Determine the geostress information of the construction area through hydraulic fracturing method to clarify the tangential geostress of the tunnel;

[0070] S2: Collect rock blocks and prepare rock samples in the area where rockburst tendency needs to be evaluated;

[0071] S3: Perform a uniaxial compression test on the specimen to obtain the uniaxial compressive strength of the specimen;

[0072] S4: Perform Brazilian disc splitting test on the specimen to obtain the tensile strength of the specimen;

[0073] S5: Calculate the in-situ stress coefficient, brittleness coefficient and elastic energy index of the surrounding rock according to the measured tangential in-situ stress, uniaxial compressive strength and tensile strength;

[0074] S6: Collect existing rockburst data and establish a rockburst case database;

[0075] S7: Using the rockburst case database, multiple machine learning models are trained to establish a rockburst intensity prediction model;

[0076] S8: Input the ground stress and surrounding rock parameters into the rock burst intensity prediction model to predict the rock burst intensity.

[0077] In some embodiments, the geostress information of the construction area is determined by hydraulic fracturing to clarify the tangential geostress of the tunnel, including:

[0078] According to the geological structure of the deep tunnel construction area, the ground stress measurement points should be reasonably arranged, including the number, location and hole depth of the measurement points;

[0079] The hydraulic fracturing method is used to determine the geostress state of the deep tunnel construction area. By drilling holes at the rock mass measuring point, sealing the two ends of the test rock layer with a packer, injecting high-pressure liquid into the sealed borehole, pressurizing until the hole wall breaks, and recording the pressure change over time. The geostress state of the deep tunnel construction area is calculated using the corresponding formula based on the fracture pressure, pump shut-off pressure and fracture orientation, and the tangential geostress σ of the tunnel is determined. θ .

[0080] In some embodiments, in an area where rock burst tendency needs to be assessed, collecting on-site rock blocks and preparing rock samples include:

[0081] Drilling sampling was carried out using a core drilling machine, and the advancement speed of the core drilling machine was controlled at 0.5mm / min~2mm / min. After drilling and coring, the cores were used to prepare 4 cylindrical rock samples with a diameter of 50mm and a height of 100mm, and 3 disc rock samples with a diameter of 50mm and a thickness of 25mm.

[0082] In some embodiments, performing a uniaxial compression test on a specimen to obtain the uniaxial compressive strength of the specimen includes:

[0083] The cylindrical rock sample was subjected to axial displacement loading at a loading rate of 0.02 mm / s using a compression testing machine until the sample was damaged, and the uniaxial compressive strength σ of the sample under uniaxial compression conditions was obtained. c and axial stress-strain curves;

[0084] Repeat the test three times, and take the average value of the uniaxial compressive strength measured by the three groups of tests as the uniaxial compressive strength σ of the rock. c .

[0085] In some embodiments, performing a Brazilian disc splitting test on a specimen to obtain the tensile strength of the specimen includes:

[0086] The disc-shaped rock specimens were radially loaded using a compression testing machine at a loading rate of 0.3-0.5 MPa / s until the specimens were damaged, and the maximum damage load P was recorded.

[0087] Formula σ for calculating rock tensile strength using Brazilian disk splitting test t =2P / (πdT), for rock tensile strength σ t Calculate, where d is the diameter of the disk and T is the thickness of the disk;

[0088] Repeat the test three times, and take the average value of the tensile strength measured in the three groups of tests as the tensile strength σ of the rock. t .

[0089] In some embodiments, the in-situ stress coefficient, brittleness coefficient and elastic energy index of the surrounding rock are calculated based on the measured tangential in-situ stress, uniaxial compressive strength and tensile strength, including:

[0090] According to the calculation formula of surrounding rock stress coefficient S, S = σ θ / σ c , calculate the geostress coefficient of surrounding rock;

[0091] According to the calculation formula of surrounding rock brittleness coefficient B, B = σ c / σ t , calculate the brittleness coefficient of surrounding rock;

[0092] The cylindrical rock specimens were subjected to axial displacement loading at a loading rate of 0.02 mm / s using a compression testing machine until the load reached 0.8 times σ c Then, the sample was axially unloaded at an unloading rate of 0.02 mm / s until the axial load was 0 MPa, and the axial stress-strain curve of the sample was recorded during the whole process;

[0093] Integrate the area enclosed under the unloading curve to obtain the elastic strain energy E e ;

[0094] The area enclosed under the loading curve is integrated to obtain the total energy of the loading process, and then the plastic strain energy E is obtained by subtracting the elastic strain energy from the total energy. p ;

[0095] According to the calculation formula of surrounding rock elastic strain energy index Wet, Wet=E e / Ep , calculate the elastic strain energy index of surrounding rock.

[0096] In some embodiments, collecting existing rock burst data and establishing a rock burst case database includes:

[0097] According to the rockburst cases reported in the existing literature, the ground stress and surrounding rock mechanical parameters in the cases are collected;

[0098] Based on the on-site observations and damage extent of rockburst cases, the intensity of rockburst is evaluated;

[0099] The ground stress, surrounding rock mechanical parameters and rock burst intensity in rock burst cases are summarized to establish a rock burst case database.

[0100] In some embodiments, a rockburst case database is used to train multiple machine learning models to establish a rockburst intensity prediction model, including:

[0101] The tangential in-situ stress, uniaxial compressive strength, tensile strength, in-situ stress coefficient, brittleness coefficient, and elastic energy index of the surrounding rock in the rockburst case database are used as input variables, and the rockburst intensity is used as the output variable;

[0102] Using the rockburst case database, machine learning models such as decision tree model, random forest model, support vector machine model, neural network model, Gaussian process regression model, and generalized additive model are trained to establish a rockburst intensity prediction model.

[0103] In some embodiments, the ground stress and surrounding rock parameters are input into a rockburst intensity prediction model to predict the rockburst intensity, including:

[0104] The tangential ground stress of the rockburst test area, the uniaxial compressive strength, tensile strength, ground stress coefficient, brittleness coefficient, and elastic energy index of the surrounding rock are input into the machine learning model;

[0105] The decision tree model, random forest model, support vector machine model, neural network model, Gaussian process regression model, generalized additive model and other machine learning models were used to predict the rockburst intensity respectively. The rockburst intensities predicted by multiple machine learning models were compared, and the prediction result with the most serious damage was taken as the final rockburst intensity prediction value.

[0106] The present application also discloses a multi-parameter rockburst prediction system based on machine learning, comprising:

[0107] An acquisition unit is used to acquire the tangential ground stress in the construction area and the uniaxial compressive strength and tensile strength of the rock samples in the construction area;

[0108] A calculation unit is used to calculate the in-situ stress coefficient, brittleness coefficient and elastic energy index of the surrounding rock according to the tangential in-situ stress, uniaxial compressive strength and tensile strength;

[0109] A model training unit is used to establish a rock burst case database based on existing rock burst data, and to train a machine learning model based on the rock burst case database to establish a rock burst intensity prediction model;

[0110] The prediction unit is used to input the tangential ground stress, the ground stress coefficient of the surrounding rock, the brittleness coefficient and the elastic energy index into the rock burst strength prediction model to predict the rock burst strength.

[0111] In some embodiments, the calculation unit calculates the in-situ stress coefficient, brittleness coefficient and elastic energy index of the surrounding rock according to the measured tangential in-situ stress, uniaxial compressive strength and tensile strength, including:

[0112] According to the calculation formula of surrounding rock stress coefficient S, S = σ θ / σ c , calculate the geostress coefficient of surrounding rock;

[0113] According to the calculation formula of surrounding rock brittleness coefficient B, B = σ c / σ t , calculate the brittleness coefficient of surrounding rock;

[0114] The cylindrical rock specimens were subjected to axial displacement loading at a loading rate of 0.02 mm / s using a compression testing machine until the load reached 0.8 times σ c , see Figure 3 Then, the sample was axially unloaded at an unloading rate of 0.02 mm / s until the axial load was 0 MPa, and the axial stress-strain curve of the sample was recorded during the whole process;

[0115] Integrate the area enclosed under the unloading curve to obtain the elastic strain energy E e ;

[0116] The area enclosed under the loading curve is integrated to obtain the total energy of the loading process, and then the plastic strain energy E is obtained by subtracting the elastic strain energy from the total energy. p ;

[0117] According to the calculation formula of surrounding rock elastic strain energy index Wet, Wet=E e / E p , calculate the elastic strain energy index of surrounding rock.

[0118] [Example]

[0119] In view of the rockburst prediction problem in the construction process of deep buried tunnels, this embodiment provides a multi-parameter rockburst prediction method based on machine learning. This method can comprehensively consider the ground stress information and the parameters such as the uniaxial compressive strength, tensile strength, brittleness index, elastic energy coefficient, etc. of the rock. The execution process is as follows: Figure 1 As shown, the specific steps include:

[0120] S1, determine the geostress information of the construction area by hydraulic fracturing method to clarify the tangential geostress of the tunnel, including:

[0121] Specifically, the implementation process of S1 is as follows: according to the geological structure of the deep tunnel construction area, the ground stress measurement points are reasonably arranged, including the number, location and hole depth of the measurement points;

[0122] The hydraulic fracturing method is used to determine the geostress state of the deep tunnel construction area. By drilling a hole at the rock mass measuring point, sealing both ends of the test rock layer with a packer, injecting high-pressure liquid into the sealed borehole, pressurizing until the hole wall ruptures, and recording the pressure change over time. The geostress state of the deep tunnel construction area is calculated using the corresponding formula based on the rupture pressure, pump shut-off pressure and rupture orientation, and the tangential geostress σθ of the tunnel is determined.

[0123] S2, in the area where the rockburst tendency needs to be assessed, collect on-site rock blocks and prepare rock samples, including:

[0124] Specifically, the implementation process of S2 is: use a core drilling machine to drill and sample, and the excavation speed of the core drilling machine is controlled at 0.5mm / min~2mm / min; after drilling and coring, use the core to prepare 4 cylindrical rock samples with a diameter of 50mm and a height of 100mm, and 3 disc rock samples with a diameter of 50mm and a thickness of 25mm.

[0125] S3, performing a uniaxial compression test on the specimen to obtain the uniaxial compressive strength of the specimen, including:

[0126] Specifically, the implementation process of S3 is as follows: a compression tester is used to perform axial displacement loading on a cylindrical rock sample at a loading rate of 0.02 mm / s until the sample is damaged, and the uniaxial compressive strength σ of the sample under uniaxial compression conditions is obtained. c and axial stress-strain curves;

[0127] Repeat the test three times, and take the average value of the uniaxial compressive strength measured by the three groups of tests as the uniaxial compressive strength σ of the rock. c .

[0128] S4, conduct Brazilian disc splitting test on the specimen to obtain the tensile strength of the specimen, including:

[0129] Specifically, the implementation process of S3 is as follows: radially load the disc-shaped rock sample with a compression testing machine at a loading rate of 0.3-0.5 MPa / s until the sample is damaged, and record the maximum damage load P;

[0130] Formula σ for calculating rock tensile strength using Brazilian disk splitting test t =2P / (πdT), for rock tensile strength σ t Calculate, where d is the diameter of the disk and T is the thickness of the disk;

[0131] Repeat the test three times, and take the average value of the tensile strength measured in the three groups of tests as the tensile strength σ of the rock. t .

[0132] S5, based on the measured tangential ground stress, uniaxial compressive strength and tensile strength, calculate the ground stress coefficient, brittleness coefficient and elastic energy index of the surrounding rock, including:

[0133] Specifically, the implementation process of S5 is as follows: according to the calculation formula of surrounding rock in-situ stress coefficient S = σ θ / σ c , calculate the geostress coefficient of surrounding rock;

[0134] According to the calculation formula of surrounding rock brittleness coefficient B, B = σ c / σ t , calculate the brittleness coefficient of surrounding rock;

[0135] The cylindrical rock specimens were subjected to axial displacement loading at a loading rate of 0.02 mm / s using a compression testing machine until the load reached 0.8 times σ c Then, the sample was axially unloaded at an unloading rate of 0.02 mm / s until the axial load was 0 MPa, and the axial stress-strain curve of the sample was recorded during the whole process;

[0136] Integrate the area enclosed under the unloading curve to obtain the elastic strain energy E e ;

[0137] The area enclosed under the loading curve is integrated to obtain the total energy of the loading process, and then the plastic strain energy E is obtained by subtracting the elastic strain energy from the total energy. p ;

[0138] According to the calculation formula of surrounding rock elastic strain energy index Wet, Wet=E e / E p , calculate the elastic strain energy index of surrounding rock.

[0139] S6, collect existing rock burst data and establish a rock burst case database, including:

[0140] Specifically, the implementation process of S6 is as follows: based on the rockburst cases reported in existing literature, the ground stress and surrounding rock mechanical parameters in the cases are collected;

[0141] According to the on-site observation and damage degree of the rockburst case, the rockburst intensity was evaluated, and the evaluation criteria are detailed in Table 1;

[0142] Table 1 Rockburst strength evaluation criteria

[0143]

[0144] The ground stress, surrounding rock mechanical parameters and rock burst intensity in rock burst cases are summarized to establish a rock burst case database.

[0145] S7, such as Figure 4 As shown in the figure, a variety of machine learning models are trained using the rockburst case database to establish a rockburst intensity prediction model, including:

[0146] Specifically, the implementation process of S7 is as follows: the tangential in-situ stress, uniaxial compressive strength, tensile strength, in-situ stress coefficient, brittleness coefficient, and elastic energy index of the surrounding rock in the rockburst case database are taken as input variables, and the rockburst intensity is taken as the output variable;

[0147] Using the rockburst case database, machine learning models such as decision tree model, random forest model, support vector machine model, neural network model, Gaussian process regression model, and generalized additive model are trained to establish a rockburst intensity prediction model.

[0148] S8, inputting the ground stress and surrounding rock parameters into the rock burst intensity prediction model to predict the rock burst intensity, including:

[0149] Specifically, the implementation process of S8 is as follows: the tangential ground stress of the rockburst test area, the uniaxial compressive strength, tensile strength, ground stress coefficient, brittleness coefficient, and elastic energy index of the surrounding rock are input into the machine learning model;

[0150] The decision tree model, random forest model, support vector machine model, neural network model, Gaussian process regression model, generalized additive model and other machine learning models were used to predict the rockburst intensity respectively. The rockburst intensities predicted by multiple machine learning models were compared, and the prediction result with the most serious damage was taken as the final rockburst intensity prediction value.

[0151] The present invention provides a multi-parameter rockburst prediction method and system based on machine learning, which relates to the technical field of deep buried tunnel construction. The method of the present invention includes: determining the geostress information of the construction area by hydraulic fracturing method to clarify the tangential geostress of the tunnel; collecting on-site rock blocks and preparing rock samples in the area where the rockburst tendency needs to be evaluated; performing a uniaxial compression test on the specimen to obtain the uniaxial compressive strength of the specimen; performing a Brazilian disc splitting test on the specimen to obtain the tensile strength of the specimen; calculating the geostress coefficient, brittleness coefficient and elastic energy index of the surrounding rock according to the measured tangential geostress, uniaxial compressive strength and tensile strength; collecting existing rockburst data and establishing a rockburst case database; using the rockburst case database, training multiple machine learning models to establish a rockburst intensity prediction model; inputting geostress and surrounding rock parameters into the rockburst intensity prediction model to predict the rockburst intensity.

[0152] The above are only preferred embodiments of the present invention and are not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A multi-parameter rockburst prediction method based on machine learning, characterized in that: The following steps are involved: Obtain the tangential ground stress in the construction area and the uniaxial compressive strength and tensile strength of rock samples in the construction area; Calculate the geostress coefficient, brittleness coefficient and elastic energy index of the surrounding rock based on the tangential geostress, uniaxial compressive strength and tensile strength; Based on the existing rockburst data, a rockburst case database is established, and the machine learning model is trained based on the rockburst case database to establish a rockburst intensity prediction model; The tangential ground stress, ground stress coefficient of surrounding rock, brittleness coefficient and elastic energy index are input into the rock burst strength prediction model to predict the rock burst strength.

2. The multi-parameter rockburst prediction method based on machine learning according to claim 1, characterized in that: The tangential geostress is determined by hydraulic fracturing to determine the geostress information of the construction area, and the tangential geostress of the tunnel is determined, specifically including: Arrange geostress measuring points according to the geological structure of the deep tunnel construction area, including the number, location and hole depth of the measuring points; The hydraulic fracturing method is used to determine the geostress state of the deep tunnel construction area. By drilling holes at the rock mass measuring point, sealing the two ends of the test rock layer with a packer, injecting high-pressure liquid into the sealed borehole, pressurizing until the hole wall breaks, and recording the pressure change over time. The geostress state of the deep tunnel construction area is calculated using the corresponding formula based on the fracture pressure, pump shut-off pressure and fracture orientation, and the tangential geostress σ of the tunnel is determined. θ .

3. The multi-parameter rockburst prediction method based on machine learning according to claim 1, characterized in that: The rock samples in the construction area are drilled and sampled using a coring machine, and the excavation speed of the coring machine is controlled at 0.5mm / min~2mm / min; after drilling and coring, the cores are used to prepare 4 cylindrical rock samples with a diameter of 50mm and a height of 100mm, and 3 disc rock samples with a diameter of 50mm and a thickness of 25mm.

4. The multi-parameter rockburst prediction method based on machine learning according to claim 1, characterized in that: The method of obtaining the uniaxial compressive strength of the rock sample comprises: The cylindrical rock sample was subjected to axial displacement loading at a loading rate of 0.02 mm / s using a compression testing machine until the sample was damaged, and the uniaxial compressive strength σ of the sample under uniaxial compression conditions was obtained. c and axial stress-strain curves; Repeat the test three times, and take the average value of the uniaxial compressive strength measured by the three groups of tests as the uniaxial compressive strength σ of the rock. c .

5. The multi-parameter rockburst prediction method based on machine learning according to claim 1, characterized in that: The tensile strength of the rock sample is obtained by a Brazilian disc splitting test, which specifically includes: radially loading the disc-shaped rock sample at a loading rate of 0.3-0.5 MPa / s using a compression testing machine until the sample is destroyed, and recording the maximum destruction load P; Formula σ for calculating rock tensile strength using Brazilian disk splitting test t =2P / (πdT), for rock tensile strength σ t Calculate, where d is the diameter of the disk and T is the thickness of the disk; Repeat the test three times, and take the average value of the tensile strength measured in the three groups of tests as the tensile strength σ of the rock. t .

6. The multi-parameter rockburst prediction method based on machine learning according to claim 1, characterized in that: The calculation of the in-situ stress coefficient, brittleness coefficient and elastic energy index of the surrounding rock according to the measured tangential in-situ stress, uniaxial compressive strength and tensile strength includes: According to the calculation formula of surrounding rock stress coefficient S, S = σ θ / σ c , calculate the geostress coefficient of surrounding rock; According to the calculation formula of surrounding rock brittleness coefficient B, B = σ c / σ t , calculate the brittleness coefficient of surrounding rock; The cylindrical rock specimens were subjected to axial displacement loading at a loading rate of 0.02 mm / s using a compression testing machine until the load reached 0.8 times σ c Then, the sample was axially unloaded at an unloading rate of 0.02 mm / s until the axial load was 0 MPa, and the axial stress-strain curve of the sample was recorded during the whole process; Integrate the area enclosed under the unloading curve to obtain the elastic strain energy E e ; The area enclosed under the loading curve is integrated to obtain the total energy of the loading process, and then the plastic strain energy E is obtained by subtracting the elastic strain energy from the total energy. p ; According to the calculation formula of surrounding rock elastic strain energy index Wet, Wet=E e / E p , calculate the elastic strain energy index of surrounding rock.

7. The multi-parameter rockburst prediction method based on machine learning according to claim 1, characterized in that: The training of the machine learning model according to the rockburst case database to establish the rockburst intensity prediction model specifically includes: The tangential in-situ stress, uniaxial compressive strength, tensile strength, in-situ stress coefficient, brittleness coefficient, and elastic energy index of the surrounding rock in the rockburst case database are used as input variables, and the rockburst intensity is used as the output variable; Using the rockburst case database, machine learning models such as decision tree model, random forest model, support vector machine model, neural network model, Gaussian process regression model, and generalized additive model are trained to establish a rockburst intensity prediction model.

8. The multi-parameter rockburst prediction method based on machine learning according to claim 7, characterized in that: The step of inputting the tangential ground stress, the ground stress coefficient, the brittleness coefficient and the elastic energy index of the surrounding rock into the rock burst strength prediction model to predict the rock burst strength specifically includes: The tangential ground stress of the rockburst test area, the uniaxial compressive strength, tensile strength, ground stress coefficient, brittleness coefficient, and elastic energy index of the surrounding rock are input into the machine learning model; The decision tree model, random forest model, support vector machine model, neural network model, Gaussian process regression model, generalized additive model and other machine learning models were used to predict the rockburst intensity respectively. The rockburst intensities predicted by multiple machine learning models were compared, and the prediction result with the most serious damage was taken as the final rockburst intensity prediction value.

9. A multi-parameter rockburst prediction system based on machine learning, characterized in that: include: An acquisition unit is used to acquire the tangential ground stress in the construction area and the uniaxial compressive strength and tensile strength of the rock samples in the construction area; A calculation unit is used to calculate the in-situ stress coefficient, brittleness coefficient and elastic energy index of the surrounding rock according to the tangential in-situ stress, uniaxial compressive strength and tensile strength; A model training unit is used to establish a rock burst case database based on existing rock burst data, and to train a machine learning model based on the rock burst case database to establish a rock burst intensity prediction model; The prediction unit is used to input the tangential ground stress, the ground stress coefficient of the surrounding rock, the brittleness coefficient and the elastic energy index into the rock burst strength prediction model to predict the rock burst strength.

10. The multi-parameter rockburst prediction system based on machine learning according to claim 9, characterized in that: The calculation of the in-situ stress coefficient, brittleness coefficient and elastic energy index of the surrounding rock according to the measured tangential in-situ stress, uniaxial compressive strength and tensile strength includes: According to the calculation formula of surrounding rock stress coefficient S, S = σ θ / σ c , calculate the geostress coefficient of surrounding rock; According to the calculation formula of surrounding rock brittleness coefficient B, B = σ c / σ t , calculate the brittleness coefficient of surrounding rock; The cylindrical rock specimens were subjected to axial displacement loading at a loading rate of 0.02 mm / s using a compression testing machine until the load reached 0.8 times σ c Then, the sample was axially unloaded at an unloading rate of 0.02 mm / s until the axial load was 0 MPa, and the axial stress-strain curve of the sample was recorded during the whole process; Integrate the area enclosed under the unloading curve to obtain the elastic strain energy E e ; The area enclosed under the loading curve is integrated to obtain the total energy of the loading process, and then the plastic strain energy E is obtained by subtracting the elastic strain energy from the total energy. p ; According to the calculation formula of surrounding rock elastic strain energy index Wet, Wet=E e / E p , calculate the elastic strain energy index of surrounding rock.

Images